There was a discussion here recently on this matter. I was waiting for a piece of work to be printed before contributing, but alas, I cannot find the thread.
I explored a line of thought which suggests a methodology of finding this 'perfect' / 'average' / 'agreed' typeface. The idea is not new: it is essentially Frutiger's 'a'. The difference here is that I used programming / automation to more quickly arrive at results. What I should've done, but didn't at the time, was to take the average left and right sidebearing's for each character also. It would be interesting to test the results on a vastly greater number of types, but of course, the method of experiment here is questionable.