The Grapheme in Classical Latin: did it represent one phoneme or two?

In Classical Latin, the grapheme <V> could be used to represent [u], [u], and [w]. There are clear minimal pairs showing that [u] and [u] were separate phonemes /u/ and /u/. However, there has been much debate about whether [u] contrasted with [w]. In one sense, it seem clear that [w] was a consonant phoneme that contrasted with other consonants: pestis /pestis/ ~ festis /festis/ ~ uestis /westis/ while [u] was a vowel phoneme that contrasted with other vowels: malum ~ malam. However, the two sounds are phonetically similar enough to qualify as allophones in complementary distribution; [w] occurring in syllable peripheries, and [u] occurring in syllable nuclei. However, the matter is complicated when we observe [w] and [u] in C_V or V_C environments.
There are four elements that I will bring into the discussion of the phonemic status of these sounds and I will conclude that [w] was an allophone of /u/. First, I will discuss conditioning elements found in C_V and V_C environments. Interesting patterns emerge where u follows /l/, /r/, /s/, and /n/. Secondly, I will discuss the evidence provided by poetry. In many cases, poetry seems to contradict the normal rules for the distribution of [u] and [w]. Thirdly, I will discuss the Claudian inversed digamma, which was used to represent [w], and what the inscriptional evidence can tell us about speakers’ perceptions of sounds. Finally, I will briefly discuss the development of [w] into the Romance languages.
This discussion hopes to provide a more nuanced understanding of the dissimilarities between popular Latin and Classical Latin, and demonstrates the role of sociolinguistic understanding historical linguistics.