I can relate to what you're saying WRT "setting it up". The parameters do do well-defined things, and their interaction is inherently logical --- but as the process works rather differently than most other processes out there, what the parameters do is also rather different, so it does take a little while to "get" what they do and how they interact. But once you do get the hang of it, it's really pretty straight-forward. One thing that does help is to read the parameter descriptions in the manual (not saying you didn't, just trying to put as many pointers as possible into this post).
Here's a quick how-to that will get you going. The procedure will of course have to be varied depending on signal, but it should give you an idea of what is what.
For removing reverb:
First, load the default setting. Set FOCUS to maximum, and set Transient Thresh all the way to the right (so it's not actually bypassing any transients). Also, set REFRACT and PRESENCE to minimum. This will result in some artifacting/bubbling for now (essentially, you're telling the algorithm to be strict and to err on the side of removing too much), but it helps tuning the other parameters by making the processing as obvious as possible.
Adjust ADAPTATION until the release slope of the visual representation of that parameter in the display matches the release slope of the input signal display. This way, you are getting a good approximation of the reverb length that you want to be removing. if you set this too low, you will hear some reverb "fading up" shortly before the onset of a new sound, sort of like compressor pumping. If you set this too high, you will start hearing parts of the sounds you wish to retain being partially removed as the algorithm will try to find reverb where it has already gone, and grabs the "next best thing".
Now activate TRANSIENT SOLO and adjust the threshold slider until you hear only the short transients coming through, but no initial reflections or sustain phase of the signal. De-activate the SOLO again. This procedure basically bypasses processing on transients to allow more drastic de-reverberation while keeping the signal crisp. Just make sure you don't let through any reflections here.
Next, slowly raise REFRACT and PRESENCE until artifacts are gone. If the signal you're working has too much reverb left when you reach the "no artifacts" range for REFRACT and PRESENCE, either try increasing LOCALIZE (which may require to also raise REFRACT and PRESENCE a little more), or try lowering LOCALIZE (which may also allow lowering REFRACT and PRESENCE).
Once you're happy, adjust FOCUS to the desired de-reverberation amount. please note that setting this to 99% instead of 100% can have a huge impact on the perceived sound....perceptively, "no reverb" is a fundamentally different experience to "nearly no reverb"! So if you find the result of full FOCUS to be somewhat artificial sounding, lowering this just slightly can potentially make it all good.
If you're not finding a sweet spot, try setting FOCUS to the center position and use the FOCUS BIAS sliders to increase or decrease FOCUS just at the frequencies where the reverb is getting in the way most. The BIAS sliders are basically offsets to the FOCUS control, frequency dependently. If FOCUS is at 50% (which corresponds to the unprocessed input signal), the sliders effectively cover the entire range from reverb-only to direct-only. As with a graphic EQ (which this isn't!!), if you lower/raise one band a lot, it may be beneficial to "smoothe the curve" by also adjusting the adjacent bands. Conversely, it can be interesting to accentuate or de-accentuate specific components of complex signals like wallas/busy backdrops by deliberately NOT using a "smooth curve" setting for the sliders.
In general, try having LOCALIZE as low as possible while still removing enough reverb (but as high as needed), REFRACT and PRESENCE as high as needed to remove unwanted artifacts (but as low as possible to remove as much reverb as possible), and ADAPTATION at a value that matches the reverb contained in your signal.
For removing direct signal:
For this application, you will want to have LOCALIZE set to maximum most of the time. Set FOCUS to minimum, set Transient Threshold all the way to the right. Set REFRACT and PRESENCE to higher values for a darker result with less signal left over, or to lower values for brighter reverb with more "foreground" components left.
For up-mixing stereo to quad:
Follow the de-reverberation strategy as outlined above for the front channels, and select "I/O Diff" for the rears...this will move whatever you're removing to the rear channels. If the absolute level of the rears is relevant (if you're anticipating fold-down to stereo that you want to sound exactly like the original stereo), make sure you do not use the transient bypass function, and set the FOCUS control to max for the front and min for the rear (and do not activate I/O Diff for the rears). Adjust the amount of transients in the front versus in the rear using the REFRACT and PRESENCE controls.
While all this just scratches the surface and you'll probably still need to experiment a little for every source, I think that by the time you've tried these suggestions you'll have some sort of feel for how the parameters interact.