Rasmussen was the most accurate in both the 2004 and 2008 Presidential elections. Almost exact predictions. Can't deny that.
That's not really accurate. Rasmussen was the closest when it came to the final presidential poll - meaning the last poll immediately preceding the election. They are shitty when it comes to state polls and not as reliable this far out from November.
Best way of looking at polls is to aggregate them and then dump clear outliers (though monitor the outliers to see if they're catching a particular sentiment early).