Alright. So let's make a tree:
bæ - pa; then
mõ - muaŋ; then
wa - mva branch; siter branch of
mpàà - mmaa.
Suppose we reconstruct the vastest order, then go up.
Firstly, let's say /à/ is just falling tone. If so, we can just assume it's no worse than not having any partic. tone on the word, so it's negligible.
One side became /mp/, the other /mm/. Let's say they came from *mb. (That might be where the tone comes from for that matter.)
We now need to solve wa - mva in order to get the two sisters. One has /w/, the other has /mv/, therefore it must have been /mwa/. (Come here and give granny a kiss. Ooh! You're so thin! *Pinches cheek* You must eat something!)
[bæ - pa]; then
[mõ - muaŋ]; then
*mwa - *mbaa.
One has /mw/, the other /mb/ - both voiced; one has a long vowel, the other does not. Maybe the long vowel has something to do with the stop - or conversely, the short vowel has something to do with the glide?
Let's go one level up: mõ - muaŋ. They share a nasal finish, a simple /m/ onset, and a nucleus that hangs around the middle back. Its cousin always has some not-/m/ onset.
Let's say it's /*muon/. Like both, but not there yet.
[bæ - pa]; then
*muon; then
*mwa - *mbaa.
There was a nasal for the lowest group, but it's gone. We need to construct an ancestor to the lowest group that will be sister to /*muon/. Maybe /*muan/, becoming /*mwan/, becoming /*mvan/, becoming /*mvã/. /*mvã/ then splits in two: /*mwã/ and /*mbaa/ (the nasality lengthening the vowel and stopping up the /v/). /*mwã/ loses nasality, either turning into /wa/, or the colouring the /w/ into a /v/. Meanwhile, /mbaa/ wither softens into /mmaa/, or the voiced /b/ produces tone - /mpàà/.
[bæ - pa]; then
*muon - *muan:
*muon: [mon ->] mõ - [muan ->] muaŋ;
*muan -> *mwan -> *mvan -> *mvã: *mwã - *mbaa:
*mwã: mvã - [*wã ->] wa;
*mbaa: mmaa - mpàà
Let's make it shorter:
[bæ - pa]; then
*muon - *muan
Let's now reconstr. the top. The basals share having a plosive to start and an a-ish vowel to finish; I imagine the distinction between /æ/ and /a/ is negligible here, and comes down to a result from voicing and having a short vowel. But that implies the original form was /*ba/.
[*ba] then
[*muon - *muan]
It doesn't make sense to say the bottom changed over time while the top did diddly-squat. Let's fig. out the bottom, then the top. Something like /*muan/ seems likely (everyone has a-es except for /mõ/) - but where did the /u/ come from, and the /n/? [*ba] has neither...
Let's say /*ba/ came from *bilabial-whatsit-presumably-voiced+/a/. Let's also assume another step happened to make... this - and that's the step between us and the protoword. Meanwhile, let's also say that /*muan/ is closer to the proto word than "this" was, but not as close as the one up the chain (say "that").
*Protoword
*"that" - something like *muan
*"this" -- *muan
*"this" --- [*muan - *muon]
*ba ----- [everything else]
Presumably, "this" was like *ba. It probably was a bilabial stop, therefore, and since most of our descendants are voiced, it probably was too. The /n/ in /*muan/ would fit nicely with that. But what about the /u/? It probably became a /w/. But then we're going out of order - the /n/ is still there, but where's the /u/? The /u/ must have disappeared first. Se need to find something that fits with /muan/, chops off the /u/ first, and then the /n/.
Say /*mwan/ -> /*mwã/ -> /*mba/ -> /*ba/ -> /pa - bæ/.
Say /*mwan/ -> /*muan/ -> /*muon - *muan/ -> [/mõ/ - /muaŋ/] - [/*mvan/ -> etc.]