Martin, thanks for your interest in my comments - and wow thanks for running some simulations to test it out! Most people would not have gone to that length. I found the floor bounce dip mitigation through modelling. Baffle width and depth and driver height and driver diameter are the parameters that I juggle. In-room measurements confirm the modelled response. One of my commercial efforts is a bipolar.
I read a two-part JBL technical paper by John Eargle once that described how to calculate the distance in a room at which the power response dominates. It was pretty complicated as I recall, involving absorption coefficients and loudspeaker directivity among other things. Here are the links:
http://www.jblpro.com/pub/manuals/pssdm_1.pdf
http://www.jblpro.com/pub/manuals/pssdm_2.pdf
I think the easier method would be to measure pink noise while walking away from the speakers. If you get to a distance where the SPL becomes steady-state, the power response is totally dominant. I doubt you would get to that distance in a typical home room. More likely is that you get to a distance where the measured SPL is +3 db compared to the predicted anechoic SPL (which falls off at 6 dB per doubling of distance from a point source). At the distance where the SPL +3 dB, the contribution of the reverberant energy (power response) and contribution of the direct sound are equal. (It's plus +3 dB instead of +6 dB because they two contributions do not add in-phase.) Anyway, in theory at least at this distance and beyond the baffle step should be somewhat mitigated.
Soongsc, a side-firing woofer (bipolar or not) would indeed avoid the baffle step for that driver. The problem is, the baffle step usually comes into play at a much higher frequency than is practical for a side-firing woofer, so it's still there for your front-facing drivers.
Another acoustic approach to the baffle step would be to use very small-diameter drivers and a very narrow baffle such that the baffle step occurs in the crossover region between the midrange driver(s) and the tweeter.
I haven't modelled this, but mounting the drivers angled back severely (like on a 30 to 45 degree slant a la Spica TC-50 or Gradient 1.3) might have a mitigating effect on the baffle step.
Duke