Hello dear troberts Thanks for your reply. I tried many things to solve this issue. One of them is lowering coefficients as you mentioned. Lowering coefficients causes real part of fitting to be slightly different from what it should be, while not fixing the imaginary part on the other hand. There is almost always a part of imaginary fitting heading toward infinity. Even if I change imaginary weight and narrow bandwidth.
I think it might be good to mention that these real and imaginary values are generated from a MATLAB code. Which I have checked validity of the results using a couple of journal papers.