Extrapolation of graphs is always invalid. And as Feynman pointed out you're usually better off ignoring the last point on the graph as well - if the last point was any good they'd have been able to make further measurements, so the last point is probably lousy.