Actually, since they're talking about variation, it can't be calculated on an individual basis. That kind of number is going to have to come from the upper bound of a confidence interval. Still, you're not going to see something anywhere near that high with genetic data on obesity. I'll stick with my idea that the 80% figure was randomly pulled otu of someones backside.
And this brings out part of the problem that many in here have refused to acknowledge. Certain people, for whatever reason, have more time than others. If you have two people with identical genetic profiles, working the same exact desk job, but one works 40 hours and the other 80 hours. Who is far more likely to be in better shape? Yeah, diet and physical activity is the key culprit, but there are other reasons factoring into why one would be more active and eat better than the other.