I have had a chance to investigate the returned speaker, so here is the post mortem report...
One discrepancy I noted between the ASR spinorama measurements and my gated measurements was the lower SPL level seen in the ASR measurements above 5.5 kHz (the approximate crossover point between woofer and tweeter). Having the reviewed speaker back, we have a pretty clear answer to that discrepancy. The yellow line is the reviewed speaker, while the red line is the speaker that stayed behind.
Response is about 5 dB lower in the reviewed speaker between 6 and 8 kHz and explains why the ASR measurements are what they are in that region.
The remaining question is whether the discrepancy is due to tweeter variation or high pass crossover component variation. Without disassembly, about all we can do is measure impedance, which should change if there is a significant difference in crossover components or a wiring error.
The low frequency differences are likely due to slight differences in cabinet leakage and port length, which were pretty quickly-cut PVC sections on a chop saw. At high frequencies, there is a 1-2 ohm difference between 2.5 kHz and 6.5 kHz, which could be impacted by the high pass filter. I'm not sure if that would be enough to produce the 5 dB difference. Probably the easiest way to isolate the contribution will be to bypass the crossover and measure the tweeters directly.
The other thing that showed up in the ASR measurements were discontinuities at 800 Hz, and two between 1 and 2 kHz. At the time I speculated that the likely causes were either port contributions that my gated measurements weren't catching or damage to the woofer. We have a gated frequency response comparison above that shows good agreement between the two speakers in the woofer's frequency range, so the next thing to do is take a closer look at the port. To do that, I placed the microphone about a quarter inch inside the port and took ungated frequency response measurements.
We clearly have our answer. The 3 primary peaks are at very similar frequencies as the ASR measurement discontinuities were and they show up in both speakers, which indicates that they weren't a result of damage to the review sample.
From this, I can conclude that the ASR spinorama measurements do agree with things I can measure with the Liberty Instruments Praxis measurement suite, an older tool, that was used by a number of commercial speaker manufacturers "back in the day". I have to keep a Windows XP PC in order to run it. The fact that I can't pick up the port contributions well in my gated measurements is something I need to think about. I also need to isolate the cause of the variation in the high frequency response, and ultimately determine how these issues could be avoided and/or improved upon for the next mini speaker project (or even if there is something "quick and dirty" that can help this design perform at a higher level).