Oh wow this is a great data set! I'll have to look into incorporating it in the future. The data I used for civilian bodyfat numbers was actually from the CDC's National Health and Nutrition Examination Survey, which used bioelectircal impedence. I know that method is pretty inaccurate but at the time it was the best data set I could find for civilians. I also figured on a population-wide basis the error factor on bioelectrical impedence would wash out a bit. Clearly this DEXA data set is better though.
As far as converting between different body fat measurement techniques I'm not aware of any good method. I think this would be difficult to do because if any of the methods were known to consistently over or under estimate bodyfat in a way that could be adjusted like that, the formulas themselves could just be adjusted to be more accurate to reality rather than converting between methods.
So the tool basically just assumes the accuracy of the number it estimates based on body measurements and also assumes the accuracy of the population-wide data sets and directly compares them. It's imperfect, but short of using the data set you've provided and asking people for DEXA results (which most won't have) I can't really think of a better way.
