I have combined your topics because they are regarding similar issues. Sorry for the slow reply, this is a difficult case and I needed to discuss it with my colleagues.
Regarding the analysis group that mimics a TFSF source, I am not convinced that this is necessary or that it would work for what you intend. If you want to measure the transmission into the surface plasmons, a square of transmission monitors around the hole should work, provided they are not placed too close to avoid measuring non-plasmon transmission.
You should note that we do not support the use of BFAST boundary conditions and nonlinear materials. Even if using them together gives the results you expect, there is no guarantee the results are reliable.
I would recommend you consider whether or not the nonlinear gain material is necessary for modelling the effect you are trying to observe. It is possible that the shift in the mode excitation is not dependent on the nonlinear gain, so you can replace the four level two electron material with an equivalent linear material with the same refractive index. You could then use an angled source and BFAST boundary conditions. You should try running the simulations with the nonlinear gain material replaced with linear material and see if you observe the expected effects.
If this does not work, you will have to perform this simulation without periodic boundary conditions in the x direction. You will have to include multiple periods and use a diffracting plane wave or Gaussian source with PML x min/x max boundary conditions. You should still be able to use standard periodic y min/y max boundaries because your source is not angled in the y direction. This approach will require a long simulation time, however. I would recommend you try the linear approach I described above first.