Kirk, you beat me to it! And it's nice to see you posting again.
A minor point, though. Wouldn't the 8 ohm tap have 3 db more voltage gain than the 4 ohm tap, not 6 db? In other words a factor of 1.414, resulting in the output transistors seeing the same load when an 8 ohm load is connected to the 8 ohm tap, as when a 4 ohm load is connected to the 4 ohm tap, as when a 2 ohm load is connected to the 2 ohm tap (since the impedance transformation will be proportional to the square of the turns ratio, and hence to the square of the voltage ratio).
Beyond that, I second Kirk's statement that "if you like what you hear, that's great." And a good takeaway from the experience reported by the OP is that experimentation with how multiple output taps are utilized can pay off big-time.
A not so good takeaway, on the other hand, would be an expectation that a similar connection arrangement will necessarily, or even probably, be optimal in the case of other amplifiers and speakers (although it certainly MIGHT be). In contrast to most amps that provide multiple output taps (i.e., tube amps), the autoformer-based McIntosh designs have output impedances that are low enough to be essentially negligible, as they are with most solid state amplifiers. And the XR100 speakers are complex four-way affairs having 15 drivers per speaker, and who knows what kind of impedance vs. frequency characteristics (I couldn't find impedance curves for them). So the OP's system is considerably different in relevant respects than the majority of others.
To the OP, again, congratulations on some imaginative experimentation that has paid off. Enjoy.
Regards,
-- Al