New CBB

The protoword of the red group could be either *tepim or *tipim. I'm going with *e because most of the other groups point to an original mid vowel.

The green group clearly has a protoword of the form *tsVnV. The first vowel is almost certainly *y, and the second one is probably *i (the rounded final vowels in the northernmost dialects might be explained by rounding harmony). It seems that vowel length is closely correlated to stress, but the diphthongization in two languages in the north whose immediate neighbours have short stressed vowels hints at a phonemic long vowel in the first syllable. This is supported by the fact that the other groups all indicate that there was once another consonant before what has become *n in Proto-Green. A subgroup in the middle of the green area has word-final stress, and all of those stressed final vowels are long, so it could even be that the true protoword was *tsyːniː.

The blue group looks fairly easy: Most languages have a *v, but some seem to have lost it. The predominant reflex of the final vowel is /a/, but some dialects have /i/ instead so I reconstruct an intermediate *e, which is actually retained in the southwesternmost dialect of this group. In the northwest, the boundaries are a bit fuzzy, it could well be that the two languages with /tɛːna/ in fact belong to blue rather than purple.

All languages in the purple group seem to have originated from a word with two consecutive medial nasals, the first of which was labial and the second of which was coronal. The medial /ə/ in two of the western dialects looks like an epenthetic vowel to me. The final vowel is again ambiguous, but the clear majority of /a/ and the absence of /i/ suggest something more open than in the blue group, which is why I'm reconstructing *a.

The orange group has clearly lost the initial /t/ of the other subgroups. The glottal fricative in the northernmost dialect suggests that there was an intermediate form with a spirant *θ, although judging only from the data from within this group, we'd have to reconstruct *h. All the orange languages have shifted to word-final stress, and the final vowel is always long. The /i/ in between the two remaining consonants in two dialects seems to be epenthetic, especially because that would explain the occasional /ɾ/ as dissimilation of *n from an immediately preceding *m. /ɲ/ in the two southwestern languages also looks like a plausible simplification of a cluster *mn.

The olive group is the most diverse subgroup. The glottal stop indicates that there must have been a plosive in the protoword, and the rounding in /tʰeʊm/ and /tʰøːm/ suggests that this plosive might have been labial (although the rounding might also have been induced the /m/ alone). Since not all languages have aspiration and none of the other subgroups have it, I suspect it's a later innovation, as is the vowel nasalisation in the northwestern dialects. Vowel length, however, is probably original because only one dialect doesn't have either a long vowel or a diphthong, but it appears even before a coda consonant in two of the languages. So, *teːpm or maybe *teːʔm.

--

For the common protolanguage, I suggest *tepne or *tebne, where either or both of the vowels might have been long. The fact that the final vowel was lost in some languages from four subgroups suggests a short vowel though; all of the long final vowels could be explained by open-syllable lengthening under stress. For the first vowel, compensatory lengthening is a plausible explanation for most of the long instances, and in those olive dialects that suggest otherwise, the length might have come from a following voiced plosive. Lastly, when compared to the forms in other branches of the family, the fact that the final nasal in the red group is labial suggests that the /i/ in the second syllable there is epenthetic.

Sound changes:

Red: ˈtebne → ˈtepne → ˈtepn̩ → ˈtepm̩ → ˈtepim
Green: ˈtebne → ˈtewne → ˈtiwni → ˈtsiwni → ˈtsyːni
Blue: ˈtebne → ˈtevne
Purple: ˈtebne → ˈtemne → 'temna
Orange: ˈtebne → ˈtemne → ˈtimni → timˈni → θimˈni → θimˈniː (→ himˈniː)
Olive: ˈtebne → teːpne → teːpn̩ → teːpm̩ (→ teːʔm̩)

It looks like there are three even more basic groups in this. Orange is almost certainly not a first-level division, but an offshoot of the purple group. Blue and green might also form a closer subfamily, and even red and olive might be more closely related to each other than to anything else despite their geographic separation:

Proto-Purple-Orange: *temne
Proto-Blue-Green: *tevne
Proto-Red-Olive: *teːpm̩