I am not convinced that Propertius II is a conflation of two separate books, as suggested by Lachmann, and more recently asserted by Skutsch ("The Second Book of Propertius", HSCP 79, 1975, 229-33), Goold (introduction to his Loeb), and Heyworth. Only two pieces of firm evidence have been offered:
1. The tres libelli: The first argument always offered by the Lachmannites is the notorious tres libelli of 2.13.25. As Skutsch puts it (230), in 2.13 "Propertius feels, or pretends that he feels, death approaching", and so gives detailed instructions for his funeral, with his tres libelli, which must then already have been written, as escort. Thus 2.13 must once have been part of Propertius' third book. (On the other hand, lines 2.24.1-2 almost certainly, and 2.3.1-2 very probably, belong in Book II, so this hypothesis requires us to imagine not just the running-together of two books, but a wholesale reshuffling.) Skutsch's argument founders on a single word. It is a tribute to Propertius' enargeía that his funeral is so vividly described that it seems to be occurring right before our eyes, but the scene (2.13.17 ff.) is introduced with the word quandocumque, 'whensoever', which clearly implies that his death is not imminent. The repeated tunc (19, 21) points the same way. What this vividly pathetic scenario means is that Propertius' love is eternal and he may as well order his funeral now, since there is no possibility of a change in plans. The number of books specified is a prediction of how many a not very prolific poet thought he might write before he died, not a statement of how many he had already written. It need not be an accurate prediction, and is not, since he wrote at least four.
2. The Grammatical Tradition: Skutsch (following B. L. Ullman) argues that the fact that Book I is not referred to in the grammatical tradition proves that it was published separately as an unnumbered 'Monobiblos', so that our Book II must represent a conflation of the original Books I and II. The analogy would be Martial's Liber Spectaculorum, Xenia ('Book XIII'), and Apophoreta ('Book XIV'), which are earlier than the book numbered I. This argument has been totally demolished by E. P. Menes ("The External Evidence for the Division of Propertius, Book 2", CP 78, 1983, 136-43), who shows that (with one or two easily-explained exceptions) none of the grammatical phenomena referred to by the grammarians happens to occur in Book I.
Since Book I had made Propertius famous, it is absurd to suppose that it was somehow omitted from the Collected Works and the Propertian Canon and survived on its own, like Martial's Liber Spectaculorum. Skutsch also argues that after Book I Propertius no longer stood for polysyllabic pentameter endings, and so (I suppose) suppressed the book as an embarrassing bit of ineptly-written Juvenilia. The problem here is that parts of Book II (e.g. 2.29) are every bit as polysyllabic as the average for Book I, and one passage (the Vergil-passage in 2.34.61-80) is considerably more so (70%).
In default of a 'smoking gun', we are left with more subjective arguments, such as the unusual length and apparent disorder of the text.
3. Unusual Length: Lachmannites always point out that there is only one parallel in Latin verse for the extraordinary length of Propertius II (1362 lines), namely Lucretius VI (1457). One secure parallel should be enough — 'einmals nimmer, zweimals immer' — but the statement, though technically true, is a bit misleading. Lucretius IV (1287 lines) and VI (1286 lines) are not much shorter, and even I-III are well above the average for Latin poetry books. Further, what little we know of Alexandrian book-rolls (surely pertinent to a poet styling himself Callimachus Romanus) shows that they were sometimes even longer: Apollonius' Argonautica runs to an average of 1459 lines per book, with Book IV, at 1781 lines, over 400 lines longer than Propertius II. (There are some hard-to-track-down hints that Callimachus' Aetia may have been on a similar scale. When it comes to collections of shorter Alexandrian poems, I don't think we have any evidence at all.) Further, we only know the lengths of six collections of shorter poems in Latin prior to Propertius II (Propertius I, Tibullus I, Vergil's Bucolica, Horace's Sermones I-II and Iambi), which is hardly enough to prove a rigid standard. (Division of Catullus into three books, I = 1-60, II = 61-64, III = 65-116, is an attractive hypothesis, but no more.) My hypothesis is that the 750-lines-per-book standard was still being established in Propertius' day, and that he could write a 1362-line book if he wanted to. Just so, he could attach -que to infinitives and other words ending in a short e, follow a verb ending in -et with the conjunction et, use oblique forms of is, and end his pentameters with words of any number of syllables, though Ovid and the Silver Latin poets could not do any of these things. In the last case, we can see the standardization occurring before our eyes from Book I of Propertius to Book IV.
Skutsch excuses the Lucretian exception by suggesting that books of didactic poetry were supposed to be longer than collections of shorter poems. Consideration of actual examples shows that post-Lucretian didactic books are either shorter than other books by the same author (Vergil's Georgics — also Horace's Ars Poetica, if published separately) or about the same size (Ovid's Ars Amatoria and Remedia Amoris), and the same pattern holds when comparing works of different authors, e.g. Manilius vs. Ovid or Martial.
4. General disorder: The fact that it is so difficult to tell where the individual elegies of Book II begin and end is no proof of conflation or gross disarrangement. If Horace's Carmina (a) had manuscripts as bad as Propertius', (b) were all in one meter, (c) were all on the same general topic, and (d) lacked named addressees who differ from one poem to the next and are generally named in or near the first line, we would be just as much in the dark about where each one begins and ends. (Note for instance the scribes' confusion at C. 1.34-35, where we have two Fortuna-odes in a row in the same meter, or at C. 3.1-6. They have much more trouble with the Sermones, where there are no metrical hints: the Epistles of course have named addressees.)
In conclusion, I see no compelling reason to believe that the textual problems of Propertius II are different in kind from those of other Latin poets, though they are certainly much worse than average in degree, perhaps even worse than those of Catullus and Lucretius.
Postscript: Skutsch and others take the steady reduction in percentage of polysyllabic pentameter-endings from Book I to Book IV as decisively refuting Gordon Williams' theory that Books I-III were published together as a unit, like Horace, Carmina I-III. The problem here is that, insofar as they are datable (most aren't), the Odes of Book I are nearly all earlier than those of Book II, which are earlier than those of Book III: cf. Nisbet-Hubbard, Vol. I, p. xxviii. If Horace could put together a unified collection while roughly preserving the order in which the parts were written, there is no reason Propertius could not have done the same. This brings us back to the idea of conflation, from a different angle. Just as Ovid conflated five books of Amores into three, a fact we only know because he tells us so in his introductory quatrain, it is possible that Propertius II was indeed made by combining two previously published and rather shorter books, but that the one who combined them was Propertius, in a second edition, perhaps inspired by Horace's Odes. Light revisions in Books I and III to make a three-book unit cannot be ruled out either, and Williams' theory has not yet been refuted.