I think what really rankles is that it fires up my developer edge-case sense. Wait, so you mean I can have one repository of 100MB, but if I have two of 50MB that costs more? Bits are bits, so why are you punishing one particular factoring of my data vs another? That's like charging for S3 per top-level directory, or for iTunes by album. I feel like charging per repository makes you feel bad about making more repositories, which can often lead to feeling bad about good decisions.
Also, realistically, a cost per developer is so miniscule compared to the overall cost of that developer. The proportional cost per project is much higher, particularly if it's some skunkworks side-project and you're the one trying to explain why you should double how much you're paying github so that you can have a live coffee consumption dashboard for the break room.
It might be irrational, but it's the truth - we associate costs together in buckets, and Github is associating itself with the wrong bucket.