Consider the phrase a month in in the following sentences:
 a. Richmond turned nineteen his third week in Vietnam. A̲l̲m̲o̲s̲t̲ ̲a̲ ̲m̲o̲n̲t̲h̲ ̲i̲n̲, he was
still looking for anything he could relate to. (source)
b. Honestly it's just been a nightmare from the beginning. My bf wanted me to quit
a̲ ̲m̲o̲n̲t̲h̲ ̲i̲n̲ without another job. (source)
How should this in be analyzed?
I have been unable to find an answer in any of my usual sources. To begin with, the OED does not record this usage of in anywhere in any of its entries for that word. Neither do any of the three comprehensive grammars I usually look at (CGEL, ComGEL, and Longman).
Absent an explicit treatment by sources known to me, two theories would seem to suggest themselves. Which, if any, is right? (This question is inspired by this question over on ELU StackExchange, and these two theories came up during discussions with others at ELU.)
Theory 1: in in  is the same as ago and on in  below
Consider the following passage from CGEL (p. 632), noting the similarity between  above and  in the passage.
A disclaimer: the thrust of this part of my question is not on whether CGEL has correctly analyzed . For example, maybe ago in  a isn't a preposition after all. My primary question, actually, is whether in in  and on in  b are both instances of the 'same thing', i.e. whether they should both be analyzed the same way, whatever way that is. (Having said that, if you think they can be analyzed in the same way, but think that ago in  cannot, I would be interested to hear why.)
Ago and on following expressions denoting periods of time
More problematic is the analysis of the bracketed phrases in:
 a. She died [ten years a̲g̲o̲].
b. [Ten years o̲n̲] nothing had changed.
Etymologically, ago derives from a past participle form agone (the prefix a- + the past participle of go), so that ten years ago has its origin in a subject-predicate construction comparable to that in [17ii] (No one — [missionaries i̲n̲c̲l̲u̲d̲e̲d̲] — had any right to intrude on their territory) except that the past participle is active rather than passive. Ago is of course no longer construed as a past participle, but how the construction should now be analysed is unclear: it is syntactically highly exceptional. Expressions like ten years occur as optional modifiers modifying prepositions (ten years b̲e̲f̲o̲r̲e̲ her death) and adverbs (ten years e̲a̲r̲l̲i̲e̲r̲), but with ago such an expression is obligatory, which indicates that it has the status of a complement. Traditional grammar classifies ago as an adverb, but on the basis that it takes a complement we analyse it as a preposition, in accordance with the criteria given in §2.4. It is even more exceptional than notwithstanding, etc., in that it always follows its complement.
On, in the sense it has in [18b], namely "later", likewise cannot occur without the accompanying phrase: *On, nothing had changed. We again take the dependent to be a complement, therefore, and treat on in this sense like ago, as a preposition that follows its complement. (One further case where a preposition follows its complement is in the idiom (all) the world over, which is essentially equivalent to the syntactically regular construction all over the world.)
Further arguments that ago is a preposition can be found in Huddleston and Pullum, p. 141:
The exceptional preposition ago
There is one preposition that is strikingly exceptional in that it invariably FOLLOWS its complement. This is ago, as in She arrived two weeks ago, where two weeks is complement of ago. The order here reflects the historical origin of ago: it derives from the form agone, containing the past participle of go. Originally two weeks ago meant something like "two weeks gone", i.e., located at a point in time that is now two weeks gone by into the past.
Dictionaries classify ago as an adverb, but there is compelling evidence that it is a preposition, exceptional only in its position relative to its complement. Consider the following data:
 i a. I spent t̲w̲o̲ ̲w̲e̲e̲k̲s̲ in Paris. b. *I spent t̲w̲o̲ ̲w̲e̲e̲k̲s̲ ̲a̲g̲o̲ in Paris.
ii a. *She arrived t̲w̲o̲ ̲w̲e̲e̲k̲s̲. b. She arrived t̲w̲o̲ ̲w̲e̲e̲k̲s̲ ̲a̲g̲o̲.
iii a. I recall his behaviour t̲w̲o̲ ̲w̲e̲e̲k̲s̲ ̲a̲g̲o̲. b. That was t̲w̲o̲ ̲w̲e̲e̲k̲s̲ ̲a̲g̲o̲.
Examples [i-ii] show that ago is head of the phrase two weeks ago: the distribution of the whole phrase is quite different from that of two weeks alone. In [iii] we see that ago phrases readily modify nouns or function as predicative complement to verbs like be: these functions are characteristic of PPs, not AdvPs, as we noted in §2.2.
Theory 2: in in  is the result of an ellipsis of an into phrase
Under this theory, the sentences in  are ellipted versions of something like
 a. Almost a̲ ̲m̲o̲n̲t̲h̲ ̲i̲n̲t̲o̲ ̲h̲i̲s̲ ̲t̲o̲u̲r̲ ̲o̲f̲ ̲d̲u̲t̲y̲, he was still looking for anything he could relate to.
b. My bf wanted me to quit a̲ ̲m̲o̲n̲t̲h̲ ̲i̲n̲t̲o̲ ̲t̲h̲i̲s̲ ̲j̲o̲b̲ without another job.
The reason why in  we have into is that the NP that follows would have been in the accusative (in the days when English had a well-developed case system). The OED explains it as follows in its entry for the preposition in:
The preposition expressing the relation of inclusion, situation, position, existence, or action, within limits of space, time, condition, circumstances, etc. In ancient times, expressing also (like Latin in) motion or direction from a point outside to one within limits; the two senses being determined by the case of the word expressing the limits, the former taking the dative (originally locative), the latter the accusative or case of direction. These cases being subsequently levelled, this distinction ceased to be practicable, and the latter relation is now ordinarily expressed by the compound in-to, into prep. and adj.; but there are various locutions in which (either because the accompanying verb conveys the sense of motion, or through the preservation of an ancient phrase without analysis) in still expresses motion from without to within.
So then the idea is that the ellipsis of expressions like those in  into those like those in  perhaps happened before the accusative got lost and got compensated by changing in into into.
Alternatively, perhaps when the ellipsis of  happens, we would be left with into on his own, which is not acceptable, and so is shortened into in (which is what it was in the first place, while NPs still all had cases).
Some reasons to think that it is not a case of ellipsis
A. It is not listed as an example of ellipsis in either ComGEL or CGEL, even though they list a lot of types of ellipsis.
B. It really doesn't fit all that well the criteria for ellipsis given in ComGEL (pp. 883-888). Let's consider some of these criteria:
(a) The ellipted words are precisely recoverable This means that in a context where no ambiguity of reference arises, there is no doubt as to what words are to be supplied: She can't sing tonight, so she won't.
This is not satisfied in my case—it is in fact not completely clear what words are omitted (is it really 'his tour of duty'? Are we sure it isn't e.g. 'his service'?).
(b) The elliptical construction is grammatically 'defective' Typically, ellipsis is postulated in order to explain why some normally obligatory element of a grammatical sentence is lacking. If such 'gaps' did not occur, there would be no obvious grammatical motive for invoking the concept of ellipsis in the first place. For example, in She can't sing tonight, so she won't, the auxiliary won't occurs without a following main verb.'
From this it follows that the burden of proof is on those who claim that something is an instance of ellipsis. And in this case, we do have an alternative possibility, namely Theory 1, which (maybe) can explain  without resorting to postulating an ellipsis.
(c) The insertion of the missing words results in a grammatical sentence (with the same meaning as the original sentence
As I mentioned, our case will fail to satisfy this unless we change in into into. But that already means we are not dealing with a typical ellipsis.
(d) The missing words are textually recoverable
From the discussion of (a) above, it is clear that our sentences do not satisfy (d), either.
But doesn't the above then show that Theory 2 is definitely wrong?
Well, not quite. ComGEL also says the following:
Ellipsis defined in terms of gradience In view of the variable application of the above criteria (a-e) for ellipsis, it is reasonable to use the term 'ellipsis' in a way which acknowledges different degrees of strictness in its interpretation.
One reason to why in might be different from ago and on is that the phrases in  could not have come about as ellipses of anything.
On the other hand, it seems that one can repeat everything in  above for in (literally just replace ago by in). True, all that shows is that in is a preposition in that phrase; but if ago and on can function like this, why couldn't in? Why postulate an ellipsis?
So, to repeat: how to analyze the in in ? Is either of my theories correct?