This is what people don't understand about cherry-picking data; with many things that don't have a direct and comparable representation (ie you get to choose what measurements make up the final results) there is almost always a combination of empirical measurements directly suited to the results you want to achieve.