That's a bummer. I think the fellows were on track with solving your issues. Well one of them, at least. Square room would still inflict it's damage. Your issue seems to occur at the crossover freq, which quite possibly could have been solved. That deep notch you measured indeed looks like a cancellation due to phase/polarity.
Consider an ideal 4th order crossover (slopes of 240db/oct, responses cross at -3db), or 2nd order (12/12, responses cross at -6db). If the levels, slopes, and polarity of the sub/mains crossover are correct, the response will sum flat at the crossover frequency. If the levels and slopes are correct, but polarity reversed, the crossover point will exhibit a deep, narrow notch at the crossover frequency. That looks suspiciously like what your measurement showed.
If you have REW (and it's helpful if your subs are in close proximity to your mains...in situ measurements can be a real mess), you can employ this phenomenon to empirically test your crossover settings.