I don't understand. Those margin of error numbers you cited is typical of any poll. The fact that they had 1001 registered voters mans that their sample size was larger than a lot of national polls.
Like I said, one broken-glass base voter is much better than two potential swing-voters (who will probably not vote in any case). And that poll was designed to mask that (just like I pointed out even before you posted the poll). Show me a poll that will show hard-core Republican voters (especially one taken of likely and assured voters) signing on and you might have a point. But conservatives will flee from a McInsane ticket (as they should)...