Cool, this means hmatrix is still an option for my fitting problem. The problem is a bit annoying but I can work with it. My assumption about the cause of the problem (wrong number of arguments) made me miss the real problem. Thanks for the explanation.