"What is the accuracy of these generic polls in elections past, as opposed to individual state Polls?"
The accuracy is: If Democrats lead by 5-8 points, Republicans have a good chance to win, as was in 2004.
Check out that poll for the 2004 Congressional election. It proves your point; look at the lead the Dems had even close to the election. That they're down to 9 points in August must be making them shake in their boots.
I just checked a few sites that trumpeted "gloomy prediction for GOP in November" stories not long ago--funny, none of them are posting THIS story...