As I never wrote an actual feature request about it: here it is !
There was the old discussion on the mailing list that the results diagram could be displayed in a "gaussian way" to avoid ceiling and bottom effects.
For this purpose the result score would have to undergo a "Z-transformation" and then the Z-score should be looked up in a "percentile table".
z= (score - average) / standard-deviation
percentile example: http://www.psychology.eku.edu/Links/prtable.htm
In this way all results would also be differentiated in a better way around the average.