Sorry you're right, on that aspect I was mistaken. But the caveats listed in the thread and the weakness of the assumption of a uniform distribution in each bucket makes it difficult to draw any meaningful conclusions beyond what the original article had. That is, I don't think the mean is useful at any level.