CST AR-Filter Farfield Problem

I'm having an issue with the online AR-Filter in CST - it can produce accurate S11 and speed up simulation greatly, but the radiation efficiency seems to be very low. Every simulation ends up with very low gain/realized gain when I use the AR-Filter. When I simulate without it, there are ripples in the S11 and the "energy state criterion" might show up, but the radiation efficiency is 2x or higher than that of the simulation with the AR-Filter. I want to do optimizations of closely spaced, electrically small antenna structures with the AR-Filter so that they don't take weeks, but I don't know if I can trust the farfield results.

Can anyone offer any tips or advice?