Different results by meshing condition in CST

Hi everyone. I am trying to simulate coupling problem using CST microwave studio.
I have some meshing problem. I know that increasing 'lower mesh limit' is helpful for the accuracy of the simulation. However, when I increase the 'Lower mesh limit' the coupling is decreased as much more than that I expect.
So, I want to know what is good meshing setup. How to determine the proper 'Lower mesh limit' ?

In most cases, then your expectation is wrong and the EM result is correct. With denser mesh, you should get even more accurate answers, the only limitation is simulation time and memory requirement.

Thank you. Yes it's true. but I have measurement result and HFSS result also. The two results are agree well but only CST results is different. I think that some setting in CST is wrong. But I don't know that....
Thank you again.

I would recommend for tetrahedral mesh in frequency domain analysis, with adaptive mesh refinement clicked on. The default value works fine , however making changes from there, should not have much deviation from what is expected.