Are they using a likely voter model at all? The survey writeup talked about a sample of "registered voters". Aren't they just calling people up, weighting by demographics (but not party ID) to get a registered voter sample, and then they get what they get? I don't think they make any assumptions about party ID per se.
Yes, they only poll registered voters (likely voters are only polled a few months before an election).
I don't really know if they weight by party or not, even though in their demographic summary it says "weighted by party":
http://www.quinnipiac.edu/images/polling/oh/oh06252013_demos.pdf