While we're all waiting for PE23, I've got a sort of (possibly, defective) idea about the pronouns. I've been studying the PE21 carefully for noun information, trying to get as much pronoun-relevant ideas as possible. And this place gets me stumbled, for it can point to a very different Sindarin pattern. It is about the three original forms of noun (p. 75):

This order was normal in Eldarin and was primitively the chief means of distinguishing what we should call 'direct' and 'indirect' objects. Thus in unemphatic pronouns (which are archaic in form and largely escaped the later inflexional elaborations), where two such occurred in a sentence, the one nearer to the verb (or most closely agglutinated to it) was taken as the direct or nearer object; the second was in function usually what we should describe as 'dative'. There was in Eldarin no distinction felt or marked between "I taught K. music" and "I gave K. a gift." In such cases in Eldarin, and some of the derived tongues, it remained possible to express both by uninflected forms.

Both by uninflected forms! In Quenya that doesn't happen — where then? In my understanding of the thought, in the unambiguous cases, such a single direct object of a transitive verb, or the single indirect of an intransitive, and even when simultaneous (expressed by word order) the pronouns, being an ancient element, remained uninflected.
Now a conjecture: there are no inflected accusatives in Sindarin pronouns at all. Le linnathon, le nallon are a normal way of expressing a single unambiguous object. In Im Narvi hain echant, the word hain is rather a plural of san < sa-na, cognate of Q. sana 'the aforementioned (one), that one previously mentioned' in predicative use, which wonderfully fits both to PE17 gloss and the Nieninkwe gloss. Only when the verb takes both types of pronouns but only an indirect is shown, an inflexion must be used, hence guren bêd enni (peded can take direct ones as well!). The problem here is what to make of 'A tiro nin, Fanuilos'. Possibly, the meaning was felt as a dative, 'watch to me/over me'; and nin is actually 'nin for a full form annin.

Or it could alternatively be an emphatic form. If 'im' is a marker of 'self' independently of person (Im Elrond echanthel), then we know no emphatic forms, but guess their existence. In PE21 also: the recipient was denoted by an unemphatic pronoun (always in Eldarin kept as close as possible to the verb): thus / gave him a horse. But if the pronoun was emphatic the normal order was maintained: J gave a horse (to) him, not you.' Look at that: the postposition of the recipient points in CE to emphatic usage. If in S. the same, then 'A tiro nin' = 'O watch mé (not somebody else)'. In Quenya, emphatics are formed with full inflexions over a vowel, inye. In Sindarin, possibly, the derivation could rather be as in Quenya expressed possessives, with an inflexion over the original stem, ni-ne > nin?