I went through almost the exact same process & steps (it’s kinda scary how identical everything you’ve mentioned is to the things I was trying - almost seems like you must have some sort of parallel universe brain scanner because you might as well have been writing from my memories. Seriously, near identical thoughts, trials, things that are holding you up/trying to figure out) a couple of years ago when I had a 3800.
In the end I found the greatest problem I had was not being able to limit the individual ink/channels as I too was using the Epson driver, couldn’t afford a rip at the time - still can’t. I found that as I would increase the ink density, 1 or 2 inks would be near perfect while others would obviously be over- or under-inking & because of the lack of granular control over what the printer is doing & how it’s laying down the ink, I couldn’t really go much further.
What I found after literally months of tests & measurements & Colorthink analyisis & PatchTool analysis & of course, prints of actual real world images (not just PDI test print images, etc) was that no matter how many test one runs or charts are measured, with the Epson driver & having only a single ink density control to really play with, other than paper types & platen gap - and platen doesn’t affect colour at all I found, just head strikes & perhaps some might say detail - it pretty much does come down to a visual inspection as Pat mentioned. I was constantly on the lookout for the graininess & break-up that occurs when there is too much ink & trying to keep that to a minimum while still having the maximum ink density setting I deemed satisfactory.
And after all those months of testing & way too many late nights, I found that I did actually prefer MonacoProfiler profiles over the PMP5 profiles. I only print RGB images, never CMYK & never proofing another device. I found that profiles from both apps would give excellent soft proof on my Eizo CG242W but I preferred the printed output from MonacoProfiler.
That made the descision about target patch numbers irrelevant as 1728 is the maximum number of patches that can be used for creating an RGB profile with MonacoProfiler 4.8. And they were definitely the equal or better than profiles created with up to ~4200 patches (that was the most my arm could take, scanning using the i1 white backboard ruler device).
As Pat again pointed out, these days it isn’t really necessary to have so many patches in our targets because the printers are so much better behaved (read linear) that around 1500-2000 was the maximum I ever needed. With the 3800, more patches didn’t seem to have any effect, good or bad. I’d have to guess that the 9900 would be even better in regard to linearity & predicatbility than 3800/3880.
Now that I have a 3880 for the past (can’t remember how long it’s been out, got it few days after release) however long it’s been, I sorted out the ink density by going for Dmax & smooth patches in my printed targets & smooth gradients or solid colour areas in images. The only problem now is that with the controls inside MonacoProfiler, allowing control over the amount of gamut compression for both lightness & chroma I’m almost creating new profiles on a per-image basis. Not quite, but it seems that way at times.
I guess it all depends what one is doing with their printed output. I wasn’t proofing or doing commercial work so deltaE numbers really only mattered to my ego. Much more important is how the final print appeared visually because that’s all that matters once it’s in a gallery or on someone’s wall.