Why Kos prediction looks wrong

I wrote a couple of days ago that an attempt to extrapolate the result of this year's U.S. presidential election "overstates the Democratic position."
Here's why.

My radar for possible bias comes from the Hillary Clinton versus John McCain match-up.
Poblano's list of variables is comprehensive:
1. John Kerry vote margin, 2004;
2. Fundraising;
3. Percentage of Baptists, especially Southern Baptists (Obama only);
4. LDS (Mormon) voters.

Then he writes:

However, the work is not done, as there are at least three other ways in which the polling data may differ from the eventual results.

These are: Sampling error, State-specific movement and National movement.

One problem is that the mechanisms to cancel out these effects run the risk of introducing their own distortions.
But what gets me thinking this analysis is a little too convenient is the state tally.
Florida, with a 3.2% Republican majority, is the closest to being lost by McCain to give him his 299 electoral college votes (remembering that Hillary Clinton get 239). But of the 50 states plus Washington D.C., 10 have smaller majorities, and ALL are estimated to go to Senator Clinton. Which seems a little unlikely.
For the sake of comparison, in 2004, there were SEVEN states in the less than 3.2% majority, and of these, John Kerry won THREE (Wisconsin, New Hampshire and Pennsylvania) while George W. Bush won FOUR (Iowa, New Mexico, Ohio and Nevada). Poblano also makes the mistake of listing Iowa as Democrat in 2004, when of course it was one of the two gains by President Bush.
It seems plausible that within the margin of error of this forecast, Senator Clinton could end up losing the following states (EC votes in brackets) in order of likelyhood (most to least): Missouri (11), Ohio (20), Washington (11), Michigan (17), Pennsylvania (21), New Hampshire (4), Minnesota (10), Delaware (3), Hawaii (4) and Maine (4, although more likely a 3-1 split). This would trigger a 403-135 or 404-134 thrashing for Senator McCain over his Democratic opponent.
To me, an interesting element of this forecast, is how well McCain appears to do, even against the Obama juggernaut. Obama picks up five states, but comes close elsewhere. Even a tidal wave for "change" does not remotely, at this stage, look like a Ronald Reagan triumph in 1980 in reverse.
If the projection that Barack Obama can't win Ohio by 1% is true, how will that square with the "Bush stole Ohio" nonsense last time round, when the majority for the Republicans was 2.5%? Silly me, only Republicans steal elections...