Sorry for the wait.
Based on your figures and after checking your simulation file, it looks like the PML boundaries were too close to device. As a result you are getting reflection from PMLs where their size can vary with wavelength. Generally we recommend everyone to do convergence testing to make sure that the results are stable with changes in the mesh size, FDTD simulation region, and number of PMLs.
I run a few simulations and looks like PECs create some plasmonic effects and the trapped light takes a long time to be dissipated. On the other hand, since PECs are very thin compared to injected light wavelength, simulations take a long time to finish.
Thanks to @fgomez, an alternative approach would be to use 2D PECs. This should be a very good approach to decrease simulation times significantly. Although in some cases we still need to use fine meshes along the material thickness (for 2D material), you can start with no mesh override. Please adjust the PECs and let me know if you encountered any problem.
I have the modified simulation file attached for your future references (since you mentioned that you are new to Lumerical), but please note that you will need to increase the simulation region, adjust the PMLs, and use finer mesh to make sure that your results converge. Generally speaking, these type of simulations that involve plasmonic effects are challenging, and for your case, convergence tests take a long time due to mandatory use of fine mesh resonator1_modified.fsp (254.7 KB).