I believe that this occurs because they set out to make a certain number of "successful" phone calls. They ask the party affiliation of each. It just happens that there were more Republicans than Democrats in those "successful" calls.
Since the registration numbers are nearly equal between D's and R's, one would expect closer numbers in a random sample. They do not intentionally set out to oversample one party over another.
I'm not a statistician but I would assume that there is some reasonable way to "normalize" the results randomly (a la Zogby's "special sauce"). Whether or not they did this, I don't know.
Though the race could hardly be statistically closer, a new poll by Zogby International indicates a likely convention "bounce" for the Republican ticket. The poll of likely voters taken during the GOP gathering in New York gives Bush-Cheney a two-point lead over Kerry-Edwards (46-44).
http://www.csmonitor.com/earlyed/early_usa0903.htm
So you believe that the limit was successful calls? I'll accept that, but do you think that it is a cost issue? Again I'm no statistician either but that seems to be a strange limit. However, I understand that there many be a valid reason for this criteria.