thick substrate simulation in ads

I 'm simulating a passive circuit(microstrip filter) using agilent ads with MOM engine simulator.
My problem is that by choice of a substarte (ε =10) thickness of 2.54 mm the simulation result goes wrong. On the other hand, s21 is above zero dB(for a passive circuit).

I should mention that by choosing the same substarte with thickness of 0.812 mm, the result is ok.

What's wrong?

Any comment is welcome

This is most likely a problem with port calibration. Can you post more information about your simulation model?

Dear volker_muehlhaus

My port configuration is shown in the following Fig.

S-param:

Filter layout ( a sample layout which is different from the the original layout but with the same substrate ):

Regards,

Amir

Try to increase the number of meshes and activate "edge mesh". Are you sure with TML-calibration?

Regards