Well, 20 consonants is a lot more than 13, let alone 10!
Looking at WALS, everything in and around southern california has an "average" consonant inventory - 19 to 25 consonants. In coastal northern california, they edge up to moderately large (26-33). There are some moderately small inventories (15-18) in inland california and in Mexico. But we're talking here about a polynesian language with a small inventory - the closest small inventory to southern california is Comanche in Texas, and after that you have go as far afield as Alabama (in Alabama) and Totontepec Mixe (in southern Mexico) to find anything similar.
These tendencies obviously aren't absolute (Mexico has one very large inventory in an area that's mostly small-to-medium), and of course WALS doesn't include every language (or even describe every language correctly). Nonetheless, the tendency exists. And while it's true that Polynesian languages progressively lose consonants as you go east, I still think it would be particularly remarkable for, as DV suggests, a language with a small inventory to arrive in an area with average inventories and react by getting an even smaller inventory, to the point of having one of the smallest inventories of any language in the world!
(a tiny vowel inventory, on the other hand, would seem to fit in perfectly).
And after all, what's the point of making a "Californian" Polynesian that doesn't take on Californian traits?
The big one of which phonologically would be syllable structure. Simple syllable structures like those in Eastern Polynesian are virtually unknown in North America, and southern california is dominated by complex structures, by WALS' definition. That is, they either allow more complex structures than CCVC, or else they allow one or both of the non-initial consonants to NOT be glides or liquids.
So yes, I'd expect some vowel loss.
And of course, complex syllable structures are associated with increasing inventory size.
So, for instance, if we take the word *takoto > tkoto, we could then have a conditioned change, > skoto.
Although I wouldn't, because I'd want some palatalisation...
So, I would probably have something like:
takoto > t'odu
tangata > tkada
huti > husi
mahina > mi:na
tika > si7a
palelau > padrag
kiko > che7u
talinga > tyega
katoa > gdo:
One thing I would question: hadn't /f/ already debuccalised to /h/ by the time of Marquesan?