It looks like whitebroncoii has bowed out with an ad hominem, but we shall motor on. Joyspring, I will attempt to address the very valid issues you have raised.

First off, my knowledge of the various filter types and their properties is limited. I have spent the most time with parallel first order passive filters as they sound the best in the ways that matter to me. I realize that no relative placement of the two drivers' acoustic centers will ever be perfect, as the path lengths vary at different locations in the listening field. For any given placement, the phase error will vary with both listening position and frequency. I do try to optimize things for the center seated listener, and go to lengths to align the acoustic centers of the drivers for the best result there. Relative placement of high and low frequency drivers' acoustic centers is best achieved through impulse measurement of each driver, and making physical adjustments until the leading edges of the impulses align when measured at the listening position. I used to do this with the DEQX unit, as its calibration made the adjustments electronically. More recently I have been using passive crossovers and aligning things by measurement with the ARTA program. Adjusting left and right channels for symmetry is aided by a stretched string tied to a button in the center of the listening chair. I used to try and eyeball things until I realized that this often resulted in errors of several inches.

I remain puzzled at the notion that high and low frequency drivers could properly reconstruct a transient when they are connected in opposite polarity... with any crossover configuration. After all, a sharp drum hit would create a steep wavefront that bears no resemblance to a periodic waveform like a sine wave. One driver moves toward the listener, the other moves away from the listener, and the result when observed with a microphone and oscilloscope is frankly chaotic. Monitoring the original event in similar fashion would yield a clean spike and some ringing after, but the recombination of drivers firing in opposite directions bears no resemblance to this in my experience.

Joyspring I cannot always determine the best polarity of a given recording for playback. The more microphones used the more confused things get, and most pop recording made in a studio are fairly hopeless in this regard. I have made a fair number of recordings using a pair of Blumlein configured figure eight ribbon mics though, and with these it is usually easy to pick the polarity that offers the best system imaging and transient response. I find it useful to place a DPDT switch at the input of passive crossovers to reverse system polarity with the flip of a switch.

I may have stepped on some toes by offering my opinion of the limitations of second order filters i.e. most stock Altecs. This was not my intent, but we learn as we go. I revere the engineers such as Harry Kimball and Jim Lansing who chose these filters long ago, and I understand their reasoning. The audio world had not yet evolved to such fussy wussy aspects as proper image localization. Jim Lansing, in his last crossover design for the late 1940s Lansing Sound two way system chose a third order network, and I have to think that he had become aware of the limitations of second order networks with their requisite polarity reversal.