I've commented once before on how Ralph Nader could actually wind up helping Barack Obama, but until Chris Bowers pointed it out, I hadn't realized how much better Obama was running in polls where the third-party candidates are called out by name. According to Chris's data, Obama is polling fully two-and-a-half points better in national polls that have run versions both with and without Nader and Bob Barr -- and the differences seem to be increasing rather than decreasing.
There is a fancier explanation here involving green M&M's and the Monty Hall Problem, but the the basic one is commonsensical. When a voter indicates that their first choice is Ralph Nader, we can reasonably infer that they do not like Barack Obama. These Nader votes are not coming from people on the progressive left who are trying to play it cute; 2000 taught Democrats a lesson that they won't forget until they take back the White House. Instead, they're coming from PUMAs and other left-of-center voters who find Obama unacceptable for one or another reason (if they were right-of-center voters who found Obama unacceptable, they'd just vote for McCain). We should not be surprised that, having determined that these voters consider Obama unacceptable, they'd pledge their votes to McCain if he's the only other choice.
But McCain isn't the only other choice; Nader and Barr will appear on the ballot in most states, and you can always sit out the election, undervote the Presidential race, or write somebody's (Hillary's?) name in. The general rule of thumb is that third party votes decrease as we get nearer to the election; that's why most polling -- and almost all of our state polling -- does not call out Nader and Barr. But if McCain is picking up 2 or 3 points' worth of support from aggrieved Clinton voters who say they prefer Nader when given the choice, I'm not sure how many of those votes he can expect on Election Day.