I did some further testing and found that my settings still weren't completely correct: high-passing the surrounds the way I did was effective at eliminating the 100 Hz bass that had been too loud before, but it also had the effect of rendering the frequency response of the LFE channel too uneven. When the LFE gain was set so that frequencies in the 50 Hz region were at a proper level, information around 70 Hz then became far too loud.
This is due to the slope of the crossover I used and the way the signals of the derived channels were interacting with each other. Issues of this kind do not appear in theatrical setups, due to the fact that bass management is not employed and most speakers are run full range, with the subwoofer only playing the LFE signal and nothing else. Home theaters, however, generally send the bass content of all channels to the subwoofer in addition to the LFE, and any phase and level mismatches between them will yield an end result that does not properly reflect what went in. This is especially true in this case, where surround bass cannot be completely eliminated from the LFE, and the LFE cannot be completely removed from the surrounds. Unlike in a theater, the final bass level is going to be a sum of the two together, and unfortunately my previous settings still did not balance all frequencies to the same level.
Since the channels cannot be completely isolated, I believe it would be more beneficial to use a shelf filter on the surrounds instead of a high-pass, since this will reduce the entire low end by the same amount. Therefore all frequencies in the LFE range will still be balanced relative to each other, and not at a slope. It may also be best to use linear phase filters, which will not be susceptible to phase cancellation between channels as normal filters would be.
So I still advocate a 48 dB/octave lowpass at 80 Hz for the LFE, but for the surrounds I'm now trying a very steep shelf filter which begins at 120 Hz and ends at -10 dB once it reaches 80 Hz. This probably still isn't perfect, but it seems to give a more balanced result, and I'm no longer hearing any significant problems.
The settings in Ozone to achieve this shelf filter are: 100 Hz as the center point, - 5 dB, and a Q value of 15. 'Surgical Mode' must be enabled to be able to use a filter with this level of steepness; otherwise the slope will be much more shallow. I do not know if other processors are able to use a filter of this kind, but I don't think it's critical for it to be exactly the same as what I've done, since less severe settings seemed to give reasonably similar results. Using an EQ plugin that provides a visual representation of the frequency response can be helpful for dialing in the settings.
I'm still trying to figure out exactly how to measure the LFE. When I tried making a downmix in another program, the result ended up being 10 dB louder than when Pro Tools performed the same task, so it seems that settings from one application may not translate very well to another. +17 dB still seems like a good setting in PT, but elsewhere it was only +7 to achieve the same result, so I'm a bit confused by this . . .