• Without any doubt, "Daur" and "Berhael" are accusative objects with regard to the ("imperative") verb "eglerio".
• "Berhael" is representing the lenited form of the name "Perhael" (Samwise/Sam)
• Supposing a (and) does trigger the following word's first consonant to have a lenited form, this doesn't disprove H.F.'s thesis of direct objects having soft mutation for, like soft mutation after article "i", then a-lenition in any case would have priority to the concurring rule. Yet, for the same reason, viceversa there would not be a disproval either!
• So, as H.F. has pointed out, his thesis' verification depends on whether "Daur" being a lenited form (of Taur) or not!

That's what H.F. is stating with regard to this:

"...we must assume that Daur is also a lenited form, the unmutated version being Taur. (According to LR:389 s.v. TÂ, TA3, "Noldorin"/Sindarin had an old adjective taur "lofty, noble", used in "ancient titles"; this would be a fitting honorary epithet for Frodo.)"

This is sounding pretty plausible, but so far - speculation!

So we have to state for the moment that there is no real proof in both directions.

• It is not attested whether or not "hain" ("they" - actually in acc. object position) is a lenited form, i.e. a basic form "sain" at all existing. We're inclined to believe this, yet there doesn't seem to be an evidence hard and waterproof for it.
• "mellon" ("friend", as has been proved, in acc. object position) is not a lenited form (this actually had to be "vellon" or - in ancient spelling - "mhellon"!).
• The word "minno" ("enter !") directly following "a" ("and") isn't soft mutation either! Yet, can this have validity, considering that also "minno" might be erroneous? So, that far, a situation like above with no proof in both directions.

Yet, the situation seems a bit different here, since - although not too obvious - the unlenited direct object "mellon" might intentionally be designed as a "wrong" form by Tolkien!

H.F. has touched upon this already, considering:

"One wonders if the lack of lenition was the reason why Gandalf misunderstood the inscription on the Gate of Moria: Pedo mellon a minno, "say 'friend' and enter". Gandalf, as we recall, at first thought it meant "speak, friend, and enter". Normally, mellon should presumably have been lenited as the object of pedo "speak" (*pedo vellon), but the ones who made the inscription had evidently ignored the normal lenition rules and given the word mellon in exactly the form it had to be spoken for the doors to open...
... Perhaps it was because of this Gandalf did not at first understand that mellon was the object of pedo "say, speak" and took it to be a vocative instead: "Speak, o friend!""

Though otherwise being sceptical enough, in this case we seem much closer to being convinced than H.F. :-) and going to tell you why:
JRRT, doing his work with enormous subtlety and a fascinating - and sometimes even strange! - sense for details, obviously had used this "little mistake" to convey his special story and give it a deeper ("technical" and, not at least, "moral") dimension.
We do not assume that Celebrimbor (or whoever had formed the inscription's very text) was not able to speak/write Sindarin correctly! It rather was for the reason pointed out by H.F. that - with regard to "mellon" - correct lenition had been neglected (this being the "technical" aspect):

(H.F.): "Of course, we don't know exactly how the "magic" or para-technological mechanism behind the doors worked, but it must have been some kind of artificial intelligence responding to the sound-sequence M-E-L-L-O-N only."

Tolkien's "moral" intention - in our personal opinion - is expressed in the story as such: After Gandalf's so many futile trials to cause the Moria Gate to open, he sat down pondering "entweder verzweifelt oder angestrengt" (either desperately or in hard concentration). Suddenly getting up and uttering the word "mellon" he finally succeeded and the Gate would open. "Ich hatte also doch unrecht gehabt" Gandalf said. "Und Gimli auch. Ausgerechnet Merry war auf der richtigen Fährte. ... Zu einfach für einen Gelehrten und Schriftkundigen in unserer mißtrauischen Zeit. Damals waren die Zeiten glücklicher..." (Sorry, don't have the original at hand!)
So, what in fact was it Merry did? (And we have to confess that we didn't even realize this at first sight!) - He simply had asked what was the meaning of those words, to which Gimli - with a certain arrogance - had replied that this of course was quite obvious. And now, of all people - davke ! - simple Merry had posed the right question... Without the special (or even wrong) spelling of the word "mellon", Tolkien's message would not have worked like that!

It's for this special reason that we're very inclined to assume there's not a second mistake within this short sentence, meaning that "a minno" is grammatical correct!

So - at this level of research - our conclusion is going together with H.F.'s opinion stated above. The more as "(H.F.) ... actually there is a variant of the Moria Gate inscription where the tengwar seem to read pedo mhellon instead of pedo mellon. (See J. R. R. Tolkien: Artist & Illustrator, p. 158.)..." whereas "a minno" had been left unchanged.

• "lasto beth" - (lit.: "Hear [the] word ... !") obviously is lenition, indicating that "peth" is actually standing in a direct-object position. Well, pretty convincing indeed - yet...
• ... what's up with "edro hi!"??? Since we've got an authentical translation for "hi" as "now", and it is attested that "now/here" in Sindarin is "si/sí", our deduction is obvious that "hi" is the lenited form of "si". (See also: Sam's "inspired" cry in Cirith Ungol: "... le nallon sí di-nguruthos! A tiro nin, Fanuilos!") Further, since there is no evidence for the use of this Sindarin word in the sense of "this here" or "this now" (e.g. comparable to Lojban {ti} meaning "this here" i.e. close to the speaker), it does not take up a direct-object position here! So what? What else could be the reason for soft mutation? Is there anything else left to think of than the preceding "edro" triggering lenition? So might there be no help but stating a rule that Sindarin "imperative" causes lenition of the following word? If this was to be true, also "lasto beth" is subject to it, consequently...
• ... devaluating the above deduction (although still not disproving the hypothesis).
• Yet, there seems to be help in our confusion! If there were different reasons for the two phrases triggering soft mutation. And here we're getting some further hints...

• "Noro lim..." Here we also have an "imperative" followed by a word, but this obviously being an adverb in trailing position modifying the verb "noro".

• "Mae govannen ..." In this example, the adverb "mae" is in a leading position. Yet this, at least, doesn't disprove the fact that adverbs also can be in trailing position (see above).
• So again ...

• ..."Noro lim..." could easily be changed to "Edro lim!" (Open fast!) instead of "Edro hi!" and, indeed, that seems to be the clue!
• Here, "si/hi" is an adverb of time (now) - or of locality (here) - directly following (and modifying) "edro", in the sense of the authentical translation given "...open now..." ! Now, that's leading us here to our happy conclusion:

Adverbs directly following a verb they modify undergo lenition. Hence, allowing a somewhat reliable proof that "lasto beth", obviously being a different case, with its lenition triggered according the direct-object mutation rule!

Happy conclusion? Well, there's still something else having to be considered ...

• ..."le nallon sí (!) di-nguruthos ...": This clause with the unmutated "sí" basically seems to be of the same structure as the above "Edro hi...!" (i.e. consisting of a verb with a trailing adverb of time). Is there any explanation for the two adverbs one time being lenited and the other time unmutated?

Well, one could think of that the Moria Gate Spell had been uttered by Gandalf, the wise and knowledgeable man well-taught and familiar with Elven tongues (who, BTW, for this very reason initially had not really understood the Moria Gate inscription!), whereas the Hobbit Samwise, a plain gardener lad, would have been hardly familiar with Sindarin grammatical subtleties like these! (And JRRT, a quite sensitive linguist, pensive and all wrapped up in his Elvish world of fancy, for sure would have been capable to also consider this!). Yet ...

... hadn't Sam had been "inspired" to this utterance! The book is stating that Sam's tongue would loosen up and his voice call in a language he didn't (even) know! So, can we really assume that this "inspired" cry wasn't grammatically correct Sindarin?! We don't think so. So again: Can it be due to a mistake? Maybe in - later - translating it to English?

• "Edro hi ammen!" is a phrase extended only by the object personal pronoun "for us", so far a construction simpler than "le nallon sí di-nguruthos ..." extended by "beneath the shadow of death, a more complex object expressing a locative relationship.
• "*Le nallon hi!" could be correct Sindarin as well and comparably easier from its construction, translating as "To thee now/here I cry!" (=lenited adverb of time or locality), but...
• "le nallon hi/sí di-nguruthos!" actually is ambiguous if not specified by means of grammar, i.e. in Sindarin, through mutation:

1. "le nallon hí di-nguruthos!" might translate as: "To thee now/here I cry, beneath the shadow of death!" (pay attention to the comma!) and
2. "le nallon si di-nguruthos!" instead, then should be translated as: "To thee I cry, here/now, beneath the shadow of death!"

We'd prefer to give "sí" as "here" (i.e. in a locative sense) since being part of the locative extension "di-nguruthos"

So, we're coming to the final(?) conclusion that neither of the two phrases quoted above has been composed mistakenly (hence, not devaluating each other respectively) but due to different grammatical styles and placements of emphasis.

• "... e aníra... suilannad mhellyn in..." (...he wants to greet his friends..)

This example clearly demonstrates that direct objects have to be lenited (mhellyn=vellyn), since there is no other rule to think of (e.g. singular article "i" or maybe "a-and" - or even "imperative") causing soft mutation here. Hence, actually giving evidence that direct-object lenition does exist.

• "... ar Elanor, Meril, Glorfinniel..." (...and Sunstar, Rose, Goldilocks...)

Is this contradictory to the above result, since one has to realize that all those persons names obviously are in direct-object position because still being referred to by "edregol e aníra tirad..." (... in especial he desires to see...). And it is obvious too that these names are not lenited ("Meril" theoretically can be unmutated, i.e. basic, or nasal mutation, whereas "Glorfinniel" certainly is not mutated at all! So, most probably, one is allowed to assume that all those names are basic forms. But why? Caused by what exception of the rule just recently deduced?
Could this - basic - form be triggered by "ar" (and), a conjunction introducing the sentence, and - respectively - connecting it to the different clusters of named persons? The answer must be "No!"

• Are "a" and "ar" (obviously both being conjunctives), though different in form, sharing the same meaning and grammatical function? Does "a" behave parallel to the article "i" - with "r" added when before a word beginning with the same vowel (e.g. ir ithil) or at least another vowel?

Obviously not, since the above example shows that it also precedes words with consonant initials (e.g. ar Hîr, ar Meril, ar Baravorn).
So, it is allowed to conclude that we have two independent words, maybe with the same or a similar meaning - or grammatical functions not totally congruent.

Yet, do go deeper in this issue, we will have to first analyze the following piece of text!

• "Arphent [Rían] ..." is easily recognizable as the two words "Ar *pent..." ("And [Rían] spoke...") - from: "ped-" (speak), "pêd" (he/she/it speaks), "pent" (he/she/it spoke)
• "Tuorna" is discernible as "Tuor-na", a seemingly rather ungrammatical way to express Sindarin "to Tuor", that commonly of course should be "na Duor" !

It is strikingly obvious that Tolkien had the intention of composing a pretty archaic-looking piece of text coming down to the people of The Third Age from ancient times. Tolkien's English of The Silmarillion sounding pretty old-fashioned as well (like the language of old nordic sagas or the Bible), he's trying here to give an impression of a very old Sindarin, an ancient tongue still retaining some grammatical features of an agglutinative predecessor (Eldarin or Quenya).

At the same time, he conspicuously also is imitating Hebrew, the ancient language of The Old Testament: not only did he create "ar" (a conjunction "and" - additional to "a") paralleling Hebrew "ve" with its very typical and characteristic introductory function to sentences - he also adopted a pecularity quite unique to Hebrew, namely to write "and" together with the following in one single word (e.g. veha'ares hayethah thohuvavohu)!

So, let us summarize:

• the conjunction "ar" is special and has little to do with the other Sindarin conjunction "a";
• the phonological influences of the two conjunctions, in the sense of whether triggering mutations or not, are different and independent from each other;
• the "ar"-conjunction very obviously triggers kind of nasal mutation (maybe nasal II, stop or - most probably - liquid mutation, the latter being typical for "or"-triggered mutation - so why not likewise going together with "ar"?

• "...aníra...suilannad mhellyn în phain..." (...wants to greet all his friends...):

This being an example for the - as we see it - basic rule that "grammatically induced" mutation is taking second place to mutation triggered by phonological circumstances! - (H.F.) In Sindarin, adjectives (including participles) following the noun they describe are usually lenited.

In the above example, "...all friends" should be "...mhellyn bain" - "pân" (all, sing.), "pain" (all, plur.), "bain" (lenition). But why "phain" which actually is Nasal Mutation?
This obviously being due to the fact that "pain/bain" here is preceded by "în" (his own) which apparently is triggering Nasal Mutation and at the same time edging out the above stated "grammatical" rule.

• So, one might expect that "ar" edges out direct-object leniton in favour of nasal mutation. Yet, there is no nasal mutation either! One still could argue that most of those persons names do not follow the "ar" directly. Yes, that's almost correct - all, except for one ...
• "ar Baravorn, [Ionnath dîn]" - seems to be the only part of the above sample text where "ar" is preceding a consonant initial directly. Yet, as it seems, the proper name Baravorn can be regarded as an unmutated form as well, but - at any rate - it is not one of the nasal mutations or the liquid mutation form it ought to be according the rule deduced just now. So what?

Since we cannot find any other reason left at the moment, so, plausibly, there seem to be only these - minimal - rules to think of as the culprits :

• Proper names (maybe nouns in general !) are not subject to nasal mutation triggered by "ar" (and).
• A sentence or subordinate clause opened by "ar" (and) and not having a verb of its own, does not indicate its direct object through lenition (e.g. ...[aníra tirad...], ar Meril...; ar Meril, Glorfinniel..., ar Eirien...; ar Iorhael... ).

• Normally, a direct (accusative) object pointed to by a predicate is indicated by Lenition (Soft Mutation), except for sentences or subordinate clauses beginning with "ar" (and) as far as these do not have a predicate of their own.
• The "normal" Sindarin conjunction "a" (and) - a form in it's own right and independent from "ar" - does not trigger (Soft) Mutation.

And in addition:

• "ar" (and) most probably is designed after the Hebrew model "v-" (vav); it triggers Nasal, i.e. - as it seems - Liquid Mutation, when directly standing in front of a verb (maybe, also a noun, yet obviously not a proper name!).
• Interrogatives - e.g. man (what) - although in direct-object position, obviously are not subject to direct-object lenition (e.g. Man agorech? - What did you do?). It's questionable whether or not those forms with initial "m-" are lenited already through traditional use (though they do not necessarily have to appear as direct objects, hence "built-in" lenition might be misleading!).
• Lenition "si" to "hi", as shown in the above sample text ("edro hi ammen"), initially seemed to be more than obscure! Sindarin so-called "imperative" actually doesn't trigger Soft mutation, yet, there appeared an obviously "new" rule, up to now unmentioned by the experts:
• verb-trailing adverbs do undergo soft mutation! Yet ...
• ... one has to be aware of the fact that particles like "si/sí" (now/here) also can form adverbs in a (unmutated) leading position, or - in different grammatical functions - with affect on mutation, be part of still other syntactical units.