Exit polls aren't even close to perfect and never have been. They suffer from strong selection effects in space and time - they only count who has voted so far, at that location.

In particular, the working population generally can't get to a polling place until after they leave work, pick up the kids, etc. Welfare slugs tend to vote earlier.

Pollsters do calibrate for selection effects after each election but the population is not static: people and factories move, job shifts occur, can't-miss TV shows are on at different times, etc, and things aren't quite the same four years hence.

As for the vote count anomalies: if he thinks this is a problem why not just pay for a recount?

As for the rest I understand this will be thoroughly explored at a convention in Roswell, NM soon.