This year's election has been hard to track, since a lot of pollsters seem to want to recreate the framework of 2008, when Democrats were on the rise and a tea party is something Anglophile women of a certain age did.
Thus, a number of pollsters have tended to tweak their samples to match 2008's partisan breakdown; thus, their polls seem to be on the Democratic side. Ones that assume a more neutral voter layout will have a more Republican slant.
Thus, you can shade your story by picking your preferred pollster. Republican-leaning press can play up Rasmussen, who assumes a more neutral distribution and gives results that are a few percent more pro-Romney than his MSM peers. The BBC on the other hand, seems to cite everything but Rasmussen in their poll of polls.
Whatever happened to just calling people and letting the chips fall where they may? Yes, I have studied political research methods back in the dark ages of the 1980 (a well-earned B- in that class), but the most tweaking that would be done is to adjust polls for demographic groups if the folks answering the question didn't quite mesh with black-Latino-Anglo or male-female breakdowns.
Tweaking for just the right partisan mix wasn't on radar back then. Why it has become so as of late is an interesting question that pollsters will have to answer for, especially if the MSM pollsters seem to have erred on the liberal side once the dust settles after tomorrow's elections.