First off, I disagree with multiple posters here who say that more likely/ less likely polls are meaningless. This is way too high a percentage to Simply write off as knee-jerk positive responses for one's own candidate. Even such knee-jerk responses are bad enough as it is as it reflects someone being absolutely deaf dumb and blind to the facts and world around them.
Well, they're bad, yes, but the question is about how do we interpret them. I just don't trust any poll respondent to answer honestly when asked "Does X make you more or less likely to support your candidate?" They're not going to answer honestly. They may not even be capable of answering honestly, because they might not know. They might just take in any scrap of data and figure that it makes them more likely to support their candidate of choice, even when it doesn't. Or they could just be reading the poll question as a chance to give a "symbolic" show of support to their candidate.
