Yes, this phase is again improved significantly, but is still far from the curve I expected. I still see about 600° of phase rotation in the bandpass range. I am not sure what to suggest. The different result suggests a measurement problem rather than speaker or something else in the audio system. I don't see any problem with the REW settings you posted. Has the mic been in nearly the same position; has the mic height changed?
Your repeated measurements are a good start. What are the normal XO frequencies? I would sweep a range less than that to see only the mid-woofers in their bandpass range. So if are using an XO at 100 and at 6k Hz then sweep from 200 to 3k Hz as the bandpass range. That's an octave above the lower XO and below the upper XO. I would expect the phase to be relative flat in that range. The phase chart will be less confusing by avoiding some room modes and any influence of the the upper XO.
If you don't already know what will happen, you could also measure at several distances from one of the mains. With the distances to the arrayed drivers being greater at closer distances I am not sure exactly what to expect however. It is more a learning experience for me. My guess would be that a 1/2" from the dust cap of one mid-woofer the phase would look pretty flat like a normal measurement. As we move away a little in the near field the phase may get very strange until we get closer to the LP. I would think that as the distance is increased to nearer to the LP the phase would again look more typical. I expected a floor to ceiling line array to be a smooth cylindrical wavefront once the LP distance is reasonable. I presume, if that is not the case, someone would have pointed that out. I guess if we cannot repeat measurement results at the LP this is just a waste of time - never mind.
Anyone else see what is wrong.