Depends if you want to model the peak asymmetry or explain it. If you just want to model it, you can just chuck in a bunch of convolutions and let them take up all of the shape.
I did this in a little while, you can probably do a lot better. There is also a talk from Arnt Kern somewhere where he talks about things like this.; I can't find it atm.
Have a look at this as an idea.