My first suggestion is to check is the material fitting. You can have a look at the material index and see if you also observe features similar to those in your absorption spectrum.
Second suggestion is convergence testing over mesh/PML. However, I do see that you already have relatively fine mesh and large number of PML layers.
Last thing I would suggest is to use Bloch boundaries to verify the results using BFAST.. Since BFAST may introduce errors in some circumstances, you can use sweeps with Bloch boundaries to get broadband results with angled incidence, similar to this example:
https://support.lumerical.com/hc/en-us/articles/1500006417822-Bloch-BCs-in-broadband-sweeps-over-angle-of-incidence
You are navigating away from the AIS Discovery experience