Your observation about the difficulty (if not impossibility) of choosing samples in open-primary states is very good.
Why is it we believe the polls when Santorum is in the lead, but when he’s not, we dismiss them? I heard a lot of theories about the polling in Florida that said that Romney wasn’t leading by as much as they polls were saying, and yet he won pretty handily. I hate to say it, but I think the polls may be accurate. Negative campaigning does work.