I'm basing this claim on the current data from the ever-excellent 538, plus a few other observations about the campaign.
538's most recent projection of the national vote share is Obama 51.8, McCain 46.8, or a +5 popular vote margin. I then compare with the current state-by-state projections to see what would happen if we see a uniform 5 percent swing to McCain, to make the popular vote exactly tied.
The result: Obama still wins the electoral college comfortably, picking up all the Kerry states + IA, NM, CO and VA, for an EV total of 286.
If McCain wins the popular vote by 0.4% he would pick up Colorado, but still lose the election 277-261.
Only if he gains a popular vote win of >1.2% would he pick up NM and VA and win the election.
However, there is good reason to believe that any swing towards McCain would NOT be uniform, and that Obama would do significantly better than McCain in the key battleground states. The reasons: money and organisation.
As is well known, Obama has a huge financial advantage over McCain, as kos recently highlighted with a detailed breakdown of ad spend. To take one of the most extreme examples, Virginia, Obama outspent McCain by nearly 9:1 in the most recent period. Assuming that this propaganda effort does influence the voters to some degree, you'd have to assume that this would buffer Obama from any national swing towards McCain.
But also - as 538 has illustrated graphically in its road reports - there is absolutely no comparison between the two ground games. The most recent report from Richmond, VA had the same story we've seen over and over again. Near-deserted McCain offices, buzzing Obama offices. I firmly believe that in all the battleground states this superior organisation will be worth at least +1% over the final polling averages, possibly a good deal more.
In short, if McCain gets a national 6.2% swing to take a 1.2% popular vote lead I don't think that would be sufficient to take Virginia. Obama's massive spend will probably limit the swing in VA polls and his superior GOTV operation will mean he almost certainly outperforms whatever the final VA polls tell us. So I'd estimate that McCain needs at least +2% nationally in order to win the Electoral College, and I think that's fairly conservative.
Obviously it would be far from ideal if Obama inverted the 2000 outcome and won the election despite losing the popular vote; it would diminish his mandate and his legitimacy, and obviously he would not have the kind of strength in congress we are hoping for. But I think it shows how successful his campaign has been that it's now put him in a position where he can afford to marginally lose the popular vote without suffering electoral defeat.
[Update] New Paradigm has pointed out that the 538 popular vote projection already assumes a tightening from current levels. To be precise, they think Obama is currently +6.2% ahead but this margin will shrink a little coming into the election. So in fact from CURRENT levels McCain needs a 6.2% swing to get back to even, a >7.4% UNIFORM swing to win, and probably a >8% swing to counteract Obama's superiority in the battlegrounds. In other words, the picture for him is even bleaker than I suggested.