When your results come back with one of the parameters at the constraint limit, it seems likely that it hit that limit and stayed there. You could try relaxing the constraint and seeing where it goes, unless there is some fundamental physical reason why it can't be lower than 0.43? I would also try different initial values. In that non-linear model the parameters are likely highly correlated, so you can end up with different results depending on the starting values, especially if the data is sparse.

Hi William A. Anderson and thanks for your answer.

The constraint of 0.43 is imposed by the model I'm using, so I guess it's a fixed boundary. As input value I also tried to give the model the variance instead of the EES and, with the same constraints, I actually obtained different different m for the three experiments. I thought the EES was the correct input to use, because it tells how well the model is fitting my data, but I can be wrong. Would it be more statistically/mathematically correct to use the variance instead?

for spheres the value of release exponent is more than 0.89, it indicates the non-Fickian release. If this value is 0.45 or less it indicates the Fickian diffusion.