Why high dielectric constant degrades performance of SIW slot array antenna?

In case of air filled waveguide or SIW with Er<2.5 performance of SIW slot array is good, and it is easy to predict by using simple calculations. But when Er=4, slot array is performing poor. Single radiating slot in SIW cavity still could be optimized and perform great. Although, when slots are combined in series, current disturbed very oddly and unpredictable along SIW array. Even with very thin slots with small displacement from SIW center axis. Scaling design to lower dielectric constant eliminates such problems, and best results are obtained with air filled waveguide (good with Er around 2). I suspect it is because of power is not radiated properly when high dielectric constant slot meets air?

Maybe it's not the dielectric constant per se, but the
dielectric's absorption losses (largely absent from an
air or vacuum waveguide). Do you compare these other
aspects of dielectric choices, besides Er?

I used lossless material and perfect conductor, with relatively simple FDTD program. I do not use vias (too complicated), just walls as in traditional waveguide. Maybe there is some problem inside FDTD source code and boundaries. For single slot in cavity i can obtain adequate results (rechecked idea in Sonnet Lite and have similar S11 for single slot, just frequency shifted). Also very long SIW without slots works perfectly. It is interesting to note that Sonnet Lite can be used to analyze SIW (cutoff frequency, via distances and diameter). Great tool to see if FDTD code works right on some simple models. Thank for your answer, I will search on losses idea.