Hello,
I am not sure of what happens, but not sure also of the accuracy of the measurements. Imagine that the discharges are measured in situ with an accuracy of 15%, which is a reasonable value, even on physical models... and see what are the levels obtained if you remove or add 15% in your discharge... or if you choose a prescribed output level corresponding to a discharge 15% lower... this might explain also the problems.
With best regards,
Jean-Michel Hervouet