Read other letters about this article
According to statistical theory, whoever has the higher number is probably in the lead. Being within the margin of error only means that it doesn't satisfy the 95% confidence criteria everyone uses as a default. So in that Rasmussen poll where McCain is up by 5 points, and 5 points is within the margin of error, we cannot be 95% confident that McCain is leading. However, if the margin of error were recomputed, maybe we could be 93% confident that McCain is leading.
There's a lot of peril in treating within/outside the margin of error as an on/off switch, and it can lead to major self delusion. And nobody needs self delusion.