However, every one of these systems that I have auditioned works wonders, particularly in the bass. To reconcile the conflict, I use DRC only for subwoofers - where it makes the biggest improvement - and run a (mostly) "purist" main signal path.
I agree with Kal - in theory there is no "purity" reason that DRC cannot be used throughout, however, from an acoustics perspective the ultra LF frequencies (below about 100 Hz) is the only area where DRC can effectively correct specific frequency related room modal effects. A waveform at 1000 Hz is about a foot long so any specific corrections are not going to apply to a large enough area of the room to be worth doing, and, above the ultra LF freqencies, DRC is really just an EQ adjustment, which can help correct a bright room, for example. In this case, it is always best to get appropriate speakers and acoustic room treatments first rather than make EQ adjustments. The same can be said for the ultra LF - bass traps and an ideal room that needs no correction is likely to sound much better than a system with heavy DRC. However, it is nearly impossible to achieve good acoustics down to 20 Hz so DRC becomes the only "practical" solution.
So in a sense "purity" applies in an acoustic sense - get the room as good as we can (as pure as possible) before doing anything else.
The impure part of DRC is that we are artificially adjusting BOTH the level of the primary AND that reflected signal in order to adjust the COMBINED level to be flat (our ears hear the combined signal but we also have a sense of primary versus reflected as well - although this sense is very poor or weak as you get to low bass frequencies). In reality, in an ideal world, your speakers would have a flat response and therefore you should only want to adjust the level of room reflections/modes (too strong or too weak).
Perhaps, the ultimate solution requires active canceling such as is used on noise-canceling headphones - imagine an array of active woofers with built in microphones that compare the audio in room delayed response and acoustic decay to the original signal (feedback) and which are able to cancel modal peaks and adjust your room acoustics to perform as desired. Designs such as this probably exist already (in labs) - for example Meyer speakers have a microphone in front of the woofer to reduce primary signal distortion already. The only issue is cost - in theory you could re-create the acoustics of any auditorium if you had such a system.