If I try to maximize the desirability of a response (least squares fit with four continous factors) in the Prediction Profiler the result strongly depends on the adjusted values of the factors (red lines) before maximizing. What could the reason be for that behaviour?
