If you assume that the sample size was large enough to get an accurate reading of how non-whites voted, then one of two things happened:
1) All non-whites voted somewhat similarly, which means that Clinton vastly under-performed with blacks
2) Blacks voted how we would have expected, which means that non-black, non-whites as a whole barely voted for Clinton
Let's look at Iowa's Census Quickfacts:
These are unadjusted numbers, that do not control for citizenship. Among Latinos, your citizenship population is going to be anywhere from 60-70% of the actual share of the population. Among Asians, it will be a bit higher - let's say 80% - and the rest more or less are accurately reflected. You then have to control for turnout for Latinos, which is going to be even more depressed in an election and
especially so in a caucus. In normal presidential year general elections at the national level, you effectively half the population share for Latinos. In a caucus, that number isn't going to be more than 40%. Among Asians, not much better. Let's adjust and estimate caucus make-up, which I think is still generous to non-whites:
So, referring to the entrance polls, non-whites went for Clinton 58-34. If we assume that 70% of Blacks went for Clinton (potentially a low estimate in my opinion), then we get:
(37.3*0.7) + (x) = 58
26.11 + x = 58
x = 31.89
31.89 / 62.7 = 51-43 Clinton for all non-white, non-black votersNow, this is among all non-white, non-black voters. Based on my electorate estimates above (which of course have a margin of error),
Latinos comprise roughly 28% of the non-white Iowa caucus electorate and 44% of the non-white, non-black electorate.
This means they're a considerable sum of that group that went 51-43 Clinton, but we can't say for sure exactly how they went. Asians, Native-Americans, biracial and other voters could have went overwhelmingly for Clinton, while Latinos went for Sanders. It could be the opposite. It could be that all groups within the non-white, non-black electorate voted relatively similarly.
I think, however, that it is unlikely that those first two scenarios were what happened. What I believe happened is that Clinton won each non-white, non-black group by 5-10 points, and so that's my general verdict regarding the Latino vote in Iowa. I haven't bothered to parse the precinct data yet, so my findings may or may not stand up when scrutinized as such.