49-34 among Latinos is not likely.
They have Latinos at 9% of the electorate, too. I know, subsample MoE, but that's your problem right there. Trump is only winning white voters by 12 points in this poll, and if that happens, Clinton wins handily.
Take the sample out and put Latino Decisions in , then Clinton goes up 3 points and Trump down 2. When over 40% of the people who vote Democratic are minority voters, you need to get those samples right in order to get an accurate vote share for the Democratic candidate.
Yeah, I'd trust Latino Decisions simply because they get a larger sample size of Latinos than most polls. If you swapped out their samples and put in Latino Decision's samples in their recent state polls, Arizona looks a lot more interesting, and Florida and Nevada start looking like locks for Clinton.
Advocacy groips tend not to provide reliable samples. They overestimated Democratic support compared to exit polls in 2012 and 2014