OK, what's the deal with Kett <tw b y>? The first two are from /tw kw/. The third... dunno, originally there was a vowel shift but that's weird, why would it also hit the semivowels?
Well, originally there were /j w/ and /v/, and w > H, v > w, something like that. There could be other sources. Maybe r_w? The lateral isn't terribly common in Kett, so maybe it developed later and never had a rounding contrast.
So we know there's something like this: p t tʷ c k kʷ s x m n ŋ j w
Rounding on coronals and velars is standard continental Haruan (see: Kannow, even Miar)...
p t tʷ c k kʷ
s sʷ x xʷ
m n nʷ ŋ ŋʷ
r rʷ j w
Vowels? Fuck it, /a e i u/ unless there were mergers in Insular.
There would be some kind of substrate for Insular that nuked the rounding except on the plosives. Also produced /o/ from /e/ in rounded environments. Maybe there should be more vowels in Insular - we know the islands have pretty big vowel inventories anyway. Ah, but they merge with diphthongs and are treated as diphthongs, so /au eu/ would work. (/i u/ are already not distinct from /ei eu/, at least not in... open stressed syllables?)
rʷ > wr or jʷ (two rhotics? old Proto-Kett had that)
sʷ xʷ > wf f (why not)
Nʷ > m
In reality, these would be a lot more complicated, of course. Maybe rounding would be lost in some environments.