fmincon in loop generates wrong optimizations

I tried to loop fmincon, but the results of the looped optimizations aren't right. fmincon only returns the answer of the previous run. It errors when a second optimization with another constraint function is used in the loop, too. The second set of parameters differs in number of variables in the objective function and structure of the constraints. If I debug fmincon step by step in the while-loop, I get the right answers. How can I solve this problem?