That's not to deny Americans have a problem with weight. I just think it's counter-productive to use metrics that people can easily see are problematic when trying to make a valid point. Using controversial measurements because the are easily available is lazy.
I think that margin of error is acceptable compared to the ease of use of the BMI.
The problem is that the BMI probably matches fairly well to a specific slice of the population. Maybe males 25-40. It likely loses accuracy of meaning the farther in age either direction you go, as well as the more different you are from the optimal subgroup. This is only barely noted in most cases (it's alluded to here in this article that Asian people might see problems at different BMI numbers). What about women vs men? What about the young and elderly? As a metric, BMI is poorly understood by the public, and this poor understanding is targeted and capitalized on by different studies. All it does is cloud our view of what's really going on.