I haven't done extensive research or anything, but it does seem like the gap closes when the survey includes more people. I'm hoping this indicates a strong margin of error in Trump's favor - though I don't know why it's so consistent.
I'm also hopeful that there will be a large amount of "normal" conservatives that would say 'Trump' in a poll, but are turned off by him enough to not show up to vote.