I am not sure if either of us can accurately say what is the point "they" are trying to make. The point I was trying to make is that a large number of developers who look at 37signals for technical leadership could be dangerously misled into using average as a performance metric.

Leaving aside the technical issues even from a customer point of view using average to measure performance is pernicious. There are scenarios where the average remains the same while almost half their customers are experiencing load times that have doubled. What should the report card say in that case?