In this case the horn equations - Websters - are wrong and predict some things athat are ficticiuos - like "cutoff" - doesn't happen in reality. I can't say for sure that this is the problem, but I can say for sure that anything derived from Webster's equation is wrong.

Problem is that Websters EQ should become more accurate at very LF, but if the calculations below "cuoff" are not done properly you could get the result that you see.