Hi heureka!
as DavidQD pointed out, this problem is a result of the calculation precision of the internal calculation algorithms. The configured precision for each function (factorial, sin, cos, sqrt..) is a trade-off between accurancy and server costs.
I was able to configure a much higher precision for the factorial function because of some new powerful servers. (for factorial inputs below 199!)
> (23-1)! mod 23 must be 22 ( or -1).
this should be now improved as well
Thank you very much for your feedback, I really appreciate it!
btw: a automatic image sizing feature will be availabe soon 
admin