Norman:
OK, I understand the problem.
First of all, you're getting the "Port inside PEC" error because your mesh controls are still too restrictive. Change the Global Meshing Settings to allow a smallest cell size of something like 0.02 (like 3 cells or so across your dielectric in the Z-direction). You might also need to change the Lines/wavelength and the second setting (forgot what it's called) so that you don't see the "blue blobs" in the mesh view window.
What is happening is that since your cell size is too large, it's not resolving the difference between your microstrip lines and the ground plane below.
Try setting Zmin and Zmax boundaries to "Open Add Space" and changing your lumped elements to discrete ports.
Remember that you still have discontinuities with this at the ends of your lines, where the current turns the corner for the discrete ports. For this reason, I'd favor the multi-pin ports since they don't have this discontinuity that might screw up your sensitive NEXT and FEXT measurements. If you have time, look over the multi-pin ports; they are better suited for this kind of thing since they don't introduce extra discontinuities that will ultimately lower your return loss for each line.
I know I'm beating the wrong drum here, but this problem is *much* easier to handle with a shielded MoM code. Because of your E boundaries, this problem is a nearly perfect example for Sonnet Lite, and Sonnet Lite would probably converge to an accurate result more quickly (and with a lot less work). I think you'll want to turn on mesh adaption in Microwave Studio to make sure that you are getting a fine enough mesh, especially if NEXT and FEXT values you are trying to resolve are on the order of -100 or even -70 dB. This is just a question of your professor figuring out which tool is optimal for what kind of homework problem. But I know that's not your problem.
-Max
Added after 34 minutes:
Norman:
One other comment. If you want to get NEXT and FEXT values for this coupled line pair, you had better use the default Gaussian signal type (excites all frequencies in your simulation band equally) rather than the rectangular pulse type. The rectangular pulse will show you the time domain response only. I guess if you only want to see actual NEXT or FEXT in time domain (not the usual curves over bandwidth), then you can use the rectangular excitation. However, be sure that you make your mesh fine enough to have time simulation steps small enough to resolve the time domain waveform properly.
I guess I've probably said enough already. Good luck with your studies.
--Max