Can you decode this graph they have on their site?
Looks like everything from about 22Hz to 400Hz is flat to within +/- 1.5db, for sound pressure levels of a little over 90db. The graph says nothing about distortion, though.
You raise a question I've never thought about before -- using speaker level connections through the sub to the main speakers. I've read that the impedance on the speaker level inputs is usually quite high (maybe 100K ohms) so as not to load the power amp. So what impedance does the power amp see if the sub's output speaker connections go to the main speakers?
A lot of subs that have both speaker in and speaker out terminals are providing them just to facilitate convenient connections. In those cases the two sets of terminals are simply wired together, with wires branched off internally from those points to carry the signals to the sub's amplifier (which as you indicated presumably has a very high input impedance). So the main amp would see the impedance of the main speakers, essentially directly.
A few subs, such as certain Definitive Technology models, have the speaker-level outputs high-passed (in their particular models generally at a fixed 80Hz frequency). I have no particular familiarity with those models, but it would stand to reason that the high-pass filter is passive (perhaps being as simple as a capacitor in series). In which case the main power amp would see the impedance of the main speakers at frequencies significantly above 80Hz, and would see a very high impedance at frequencies significantly below 80Hz.
However, that approach would certainly not seem like a recipe for good sound. Frequency response irregularities would be introduced if the impedance of the main speaker were not perfectly flat throughout the crossover region (which would be pretty wide if the high-pass filter were just a capacitor in series). Bass damping for the main speakers would be degraded by the increase in the source impedance that is driving it. The crossover element(s) itself would introduce its own non-idealities into the signal path, potentially affecting any and all frequencies up through the treble. And the high-pass crossover frequency would probably vary considerably from the 80Hz depending on how different the main speaker's impedance around that frequency is from 8 ohms or whatever main speaker impedance the sub's designer assumed it would be working into.