They're never really correct.
In September of 1980,
http://www.cnn.com/ALLPOLITICS/1996/analysis/back.time/9609/15/index.shtml">Reagan and Carter were tied at 39%. Reagan won by 10-points.
In September of 1992, Clinton led Bush by a
http://query.nytimes.com/gst/fullpage.html?res=9E0CE2DE143CF931A1575AC0A964958260">staggering 21-points. He won by 6.
In September of 1996, Clinton led Dole by
http://people-press.org/report/122/solid-clinton-lead-small-gain-for-congressional-democrats">18 points. He won by 9.
In September of 2004, Bush led Kerry by
http://people-press.org/report/?pageid=889">8 points nationally. He won by 3.
Now these polls are mostly different and it's impossible today to find averages, however, it shows one clear thing: The outcome of the race was far different than the polls suggested in September of each election year. In most cases, the candidate who led, won, but they saw the race tighten, except for Reagan, who rolled after the debates and easily crushed Carter.
You can't look at the polls today and get a sense of how things will be in November. A lot can change. A debate performance can wreck a candidate (as it did for Carter), a convention can get a candidate back into it (as it did for Gore) and the numbers could level out, as they managed to do for Bush and Clinton in 1992. For all the talk about Obama not living up to his potential, Clinton blew a huge 21-point lead and even though he cleaned up in the electoral college, the race took a strong turn toward Bush after September.
The same thing happened to Carter, who blew a big lead against Ford; Dukakis, who blew a big lead against Bush I and Bush II; who blew a big lead against Gore.
So what should we be looking at in these results? Not necessarily the fact the candidates are so close, but who has the more excited base. Hands down, Obama wins. Just as Bush won in 2004 and 2000 and is a big reason why he won in close states. Turnout is key and for all the polls, they can't predict who and who will not turn out. McCain supporters might say they're voting for McCain, but if they're not enthused, there is a good chance they stay home. Now this can change if he picks someone who energizes the base, but if it's a middle of the road, no flare pick, don't expect it to.
The fact is, Obama isn't winning and that isn't good. However, it isn't bad, either. It's just a middle ground between losing and wrapping this election up. But remember, even if Obama were up by 10+ points nationally, it wouldn't mean a lick...because the polls this far out are rarely ever right.