I checked the file and did some testing by changing the “tolerance 1” and “tolerance 2” settings of the EME advanced options tab based on the recommendations from the following page under the “advanced settings and how to use them” heading:
I used tolerance 1 = 0.5 and tolerance 2 = 1e-5
I also tested whether the method of sweeping the group spans to get the broadband result was valid by setting up a parameter sweep to sweep over wavelengths from 1.54 to 1.56 um over 50 points.
I found that the results stayed very similar where the peak in reflection has a much smaller amplitude for the etch depth of 75 nm compared to 50 nm and 100 nm etch depths.
Since the results don’t change with a finer mesh or increased number of modes, and the results are consistent between using the method of sweeping the wavelength and sweeping the group span, it suggests that the results could be physical. If you have access to FDTD Solutions, it might be worth doing a check between a 3D FDTD simulation and the EME results (for a fewer number of periods so that the simulation time is reasonable).