<< Is most of the knowledge base on horns from experimental evidence, or are they well understood mathematically? >>Actual horns are pretty much understood mathematically, however the behaviour of real-world speakers in these horns isn't. Result: experimentation to achieve best results.
Cheers