I think what happened, if memory serves me correctly, is that few organizations polled over the weekend before the election. Virtually every poll by the Friday before the election showed Bush with a comfortable (beyond margin of error) lead. But the DUI story broke late, and those polls didn't reflect public reaction to the DUI story. Zogby's poll did reflect public reaction, and it showed a surge (2% lead) for Gore. But by election day, the surge for Gore had crested, and the momentum was back toward Bush (hence, the dead-heat). Bush with a 5-9% lead 5 days before the election; the DUI story came out; Gore "surged" to a 2% lead by the weekend (as Zogby showed); but his surge was reversed by the end of the weekend or Monday, and voters started to swing back to Bush (the dead-heat by Tuesday). My theory is that had the election been held on Wednesday, Bush would have won by a full percentage point or two -- the momentum was in Bush's direction.
In regard to voter fraud, I've heard it said that every election has a fraud factor (intentional or accidental) of about 2%. In close elections, the fraud is magnified. But 2% -- that's two million+ votes in a presidential election! And Gore claims to have "won" the popular vote by half-a-million? Hardly. That claim, despite the "final numbers" is falacious, IMHO.
Yes, Rasmussen blew it big time. They so admitted afterwards if I remember correctly. I think they are still around but not sure if they are into the polling business now.
It seems that the DWI thing had an effect even though it was out there earlier and didn't get a response then. But the blast faxing of it the last couple of days did the trick.