I mean, I've heard of outliers, but it's anywhere from 8 to 17 points out of step with every other general election poll.

IIRC, I think I read somewhere that the LA Times poll weights the responses based upon reported 2012 vote. So if a state went 52-48 Obama and the poll respondents report 55-45 Obama, the responses of the 45% who claim/admit to having voted for Romney will be given greater weight than the responses of those who report voting for Obama to fix up the sample to line up with the known 2012 vote. Experts speculate that people who vote for losing candidates tend to misreport, thus falsely inflating the weight given to the loser's voters.