Actually the discrepancy stems from underestimating voter turnout for Trump and over estimating it for Clinton and is not caused by the mythical timid Trump voter.
Doesn't matter. All that matters is that the errors are correlated. If voter turnout is biased in one state, then it likely is in another. The reluctance was just an example.