RCP is using four polls with skewed samples:
Survey USA, uses a +7% Democrat sample to get Obama to +3%. If we unskew that data to a +3% Democrat Sample, using Survey USA's own internal data of how Republicans, Democrats and Independents are polling, we find Romney with a slight lead of 44.7% to Obama's 44.0%
CBS/Quinnipiac, uses a +9% Democrat Sample to get Obama to +5%. If we unskew that to a +3% Democrat sample, and using CBS/Quinnipiac's own data, we find Obama's lead has cut to one tenth of the poll's prediction to 0.5%: Obama 45.8% to Romney 45.3%
Gravis Marketing used a sample of +9% Democrat, and that showed a tie. If we unskew it to a +3% Democrat sample and use Gravis' own internal data,
The PPP and Fox News polls could not be accurately unskewed because they did not publish their internal data did not provide the necessary breakdowns of how Republicans, Democrats and Independents intended to vote.
Bottom line, exactly what I've been telling you is true: the RCP average, when a Drive-By Media source throws a wrench into the mix and skews the average, it spoils the entire sample. Here in reality, Romney is winning. And that's all that matters.