In a 50-state poll two or three state results should be outside the margin of error based on statistics alone.
Yep. People automatically say "junk" to these 50-state polls because of odd inconsistencies, but odd inconsistencies are to be expected simply from statistical noise. That doesn't mean all the data is worthless. It means we should be considering all polls, regardless of whether they're released as a single state poll or a group of them, with the proper amount of skepticism. They're useful, but there's bound to be errors.