Dave,
I understand about the reverse polarity in the x-over design. However, it's usually used to smooth the response and remove the dip at the crossover point. In this case, there is a huge dip at x-over, reversing polarity, or re-reversing polarity, removes the dip, smooths the overall response to where it looks like the mfgr's published response.
Measurements were taken with Smaart, and confirmed by listening to it. Actually the response curve looks like what B&C published. Still a great little box.
Don