The entrance poll has a very, very small sample size of Latino Republicans, perhaps only 130, which means that even if everything else is perfect in its methodology, it carries a +/- 8.5% points on the Latino sample. Further, the Nevada entrance polls are not designed to get accurate subgroup vote share estimates, but rather report on statewide numbers, so their design is not trying to capture a representative sample of Latino Republicans, which adds some amount of unknown bias, beyond the +/- 8.5%

David Damore, Latino Decisions