Maybe it is not clear for you when I suggest to simulate portion of the substrate. If you want to include it, it is ok. But you will need much longer simulation time to make sure the signal due to multiple reflections decays completely in the device. Currently the time monitor only recorded the first refracted singal.
You can move the monitor to a closer location;
what value do you expect the maximum in the time monitor? if the signal has been absorbed/diffracted/scattered through the way, it should not reach the source magnitude.
Maybe you need to step back, and think how this device works and what the physics it has to obey. Please make sure the time-domain simulation is complete, and in your case I guess the autoshutoff min 1e-5 is not small enough.
I still woi=uld suggest you to simulate a small portion of the substrate, and check the result to see if it is expected, such as the time signals, the reflection R and transmission T (power quantities) in frequency domain, steady state. You may refer some online edu vidoes.