Marv, what you're referring to as comb effects is more like the typical LF room response (with peaks indicating the resonance frequencies (unless the mode is poorly excited by the speaker or or the mic is in a node region for that mode hence not seeing it). The dips in between resonance is just because there's nothing else going on (the response is the sum contribution of all modes at the frequency so you get some contribution from to the resonances below and some from those above).

Looking at your graph the roll-off starts below 200Hz (the horn loading gives it the gentler slope I assume?). The driver itself has an Fs around 50Hz I guess? at least, it seems the loaded resonance is there such that the response drops sharply below that.

In any case, it's a good thing you did not try too hard to compensate for this, as I don't presume the full range driver has so much excursion freedom?

Looks like a nicely behaved driver overall though,

Humm, interesting. If you can give me the dimensions of the horn (some 2D diagram would be awesome, maybe some scan of the assembly brochure?), I can try a simple simulation of the horn, we'll see what it really does vs. a planar baffle (or no baffle at little even though it's a little silly).

Speakers driven straight out of the Schiit headphone amp 

I have no idea what the Mjolnir power output is at the nominal 8 ohms of those Fostex (rises to 60 odd something at 45-50Hz and 32 ohms at 20kHz). The Crest amp seems to provide better bass extension, solidity, and slightly better control. The Mjolnir output has more air (not unexpected giving its 1+ ohm output Z) and clearer with more microdynamics and plankton - it's more involving of a listen.

You guys didn't notice my super high-bandwidth speaker cables?