The data is hard to trust, since there are duplicate entries in the graphs. If they are not careful enough to proof their results, how I can be assured that they are careful in the way they collect...