Perhaps the answer to this seeming inconsistency has already been posted honestly a few days ago. It says that according to USA/Gallup Republicans and narrowed party identification one point shortfall, 48 to 47. This would mean that instead of 5.7% difference there is a 1% difference and the poll as therefore oversampled by 4.7%.
Allahpundit flagged an interesting point in an updated version of the USA Today story on their latest polling in the presidential race. According to the internals of the Gallup survey, McCain’s ascension to the leadership of the GOP may have healed the brand. Democrats lead Republicans in party identification by a single point:
In the new survey, more voters call themselves Republicans. Now 48% say they’re Democrats or lean to the Democratic Party; 47% say they’re Republicans or lean to the GOP.
http://www.freerepublic.com/focus/f-news/2077484/posts
This would mean that instead of 5.7% difference there is a 1% difference and the poll as therefore oversampled by 4.7%.As covered earlier this poll is using a 7.6% differential, not a 5.7. And the 1% difference is from the Gallup poll, based on random selection. So one poll using another poll's statistics for weighting purposes would be a risky proposition.