I've listened to about 30 modern Latin speakers and except for Luke Ranieri, none of them pronounce the final 'm' as it should be pronounced which is as a nasal vowel, at least during conversation. Even the greatest Latin speakers in the world such as Daniel Petterson and Jorge Tarrega do not pronounced the final m nasally. During poetry recitals the final 'm' at least is elided though I can't remember if poetry reciters pronounce m nasally when the following word begins with a consonant. Moreover, I will hear some modern speakers elide the final 'm' when it is followed by 'est' but that is the only exception. This is a contradiction. If you admit that the final 'm' should be elided when followed by 'est', then you do so not because 'est' had some strange properties but because the final 'm' was a nasal vowel and the romans elided final vowels when the following word began with a vowel.
Admittedly I do not agree with many of Sidney Allen's characterization of the vowels especially when he cites no evidence which is often but for his passage on 'm' he provides good evidence and I think he is right. Is it the case that modern speakers find it too difficult to worry about this minor detail or perhaps there are too many details and this one just slips through the cracks or is it the case that the consensus on Latin pronunciation has rejected Allen's views on the letter m?
Here is the relevant passage, excuse the typos the book was scanned:
At the beginning and in the interior of words the sound repre- sented by m presents no problem. It stands for a bilabial nasal, as in e.g. English mat or camp. There are, however, points to notice where it occurs at the ends of words. In general it seems to have been reduced (like the n before a fricative internally) to a mere nasalization of the preceding vowel—in the imprecise terminology of the grammarians it is ‘almost a foreign letter’ (+ Velius Longus, K. vii, 54), or ‘obscurum in extremitate dictionum sonat’ (Priscian, K. ii, 29); and in early inscriptions one often finds the final m omit- ted, e.g. in the third-century epitaph of L. Corn. Scipio: honc oino ploirume cosentiont... duonoro optumo fuise uiro (= hunc unum plurimi consentiunt... bonorum optimum fuisse uirum). In the course of the second century, the official spelling established the writing of final m; but forms without m continued occasionally to be found. That the vowel was lengthened as well as nasalized is suggested by the fact that such final syllables, when followed by an initial con- sonant, count as heavy—thus, for example, Italiam fat6 = hala fato. An indication of this lengthening is also perhaps seen in Cato the Elder’s writing of diem as diee (Quintilian, ix, 4, 39).26 The non-consonantal nature of final mis also shown by the fact that syllables so ending are elided in verse in the same way as if they ended in a vowel (with rare exceptions: e.g. Ennius milia militum octo: cf. p. 81, n. 3); from which one concludes that they simply ended in a nasalized vowel. For the m in this position, when followed by an initial vowel, Verrius Flaccus is said to have favoured writing a half-m (a) only (Velius Longus, K. vii, 80); Quintilian (+ ix, 4, 40) describes it as hardly pronounced; and later grammarians refer to it as being completely lost (e.g. t Velius Longus, K. vii, 54). If elision involves complete loss of the final vowel (cf. p. 78) the distinction between nasalized and non-nasalized in this context is of course purely academic. The same treatment of final mis seen in cases of ‘aphaeresis’, where inscriptions regularly omit it (e.g. scriptust for scriptum est). It is of interest that preferences regarding the elision of vowel m are the same as for long vowels or diphthongs2/—a further indication that the vowel was in fact not only nasalized but length- ened.28 xx Where, however, a final m was followed by a closely connected word beginning with a stop (plosive or nasal) consonant, it seems to have been treated rather as in the interior of a word, being as- similated to the following consonant (in this case, naturally, without lengthening of the preceding vowel). Thus we find inscriptional tan durum for tam durum (and in e.g. tam grauis we may assume a parallel assimilation to the following velar, giving [n] for m); Velius Longus says that in etiam nunc ‘plenius per n quam per m enun- tiatur’; and Cicero also refers to unfortunate doubles entendres in such phrases as cum nobis
