OK! I'll admit I didn't read every post on this thread, so if I'm repeating someone else's thinking, I beg your pardon, I never promised you a rose garden...OOPS!
Anyhoo, after 25 yrs of grappling with that very issue, what I've discovered as a drummer, who is the first one to be yelled at when the tempo is off (and the last to get the limelight) is that what makes MIDI sound like...well...MIDI is NOT the quantization, or timing in general. What makes it sound mechanical is the lack of velocity sensitivity. I even put a post on here, and Harmony Central about it years ago asking why couldn't the same council of muckety mucks that put the GM protocol together reassemble and figure out a way to put maybe 1024 increments of velocity, instead of the 128 we have now. My thinking was all they would have to do is divide that 128 (an even #) into the 1024. Both common #'s we all see in PC technology. I did get a good, knowledgeable reply from a much brainier person than I that made sense why that couldn't really happen, but I can't remember what it was.
I also had it proven to me in a much more practical way in a live performance. I went to see this band that was really a great band. In fact, when I walked in the room before I could even see the band, I was initially amazed at how clean the mix was, and even the volume was such that you could even speak to others without going horse. When I saw the band the first thing I noticed was that the drummer was playing ALL pads, INCL the snare and cymbals. He was a good drummer too. He kept the tempo just fine. However, after about 30 mins I felt like I was going to break out in hives, and I couldn't figure out why. I spent the next 30 mins trying to figure out what was making me feel this way. Finally, it dawned on me. The drum samples the drummer was using were too static, and made him sound like a drum machine with all the finesse of a Yamaha RX7; which had no velocity sensitivity.
If you're not a drummer you may not realize this, but I've tried a test several times throughout the years as MIDI, and quality sampling has evolved. If you play a real drum with simple 1/4 notes (L R L R) you'll notice there's always SOME kind of difference, no matter how hard you try to make each stroke identical to the other. When you use MIDI to do the same thing you won't hear that difference. Further, I know EXACTLY how to quantize a theoretical buzz roll with MIDI, but again, no matter what I do it always comes out sounding like an Uzi. Also, the only problem that is solved by getting a human to play a drum part is that they won't play more than 2 voices (other than the kick) at the same time with their hands, and their fills seem to make a little more sense. But that's to the very trained ear. I've owned an electronic kit before, and while it's very handy in a MIDI studio for those very reasons, it SUX to use one live, unless you're a sorry drummer that doesn't know how to tune, or mic your drums.
So, what I would concentrate on (at least as far as humanizing drums) is quantize EVERYTHING...EVERYTHING on your drum track to absolute ZERO. Then find REALLY good samples that sound like they have a lot of multisampling going on. THEN, bring in actual humans to play the other instruments, and THAT will give you the humanization you're looking for. I PROMISE!
For what it's worth, SampleTank uses some very well-sampled drums. Go listen to some really good session drummers like Steve Gadd, or the late great Jeff Porcaro. Listen to all of the subtlety they use especially on the snare. The other drums you can get away with not so good samples. Another example is how Alesis drum samples usually are KILLER sounds just by themselves, but using them in a fast roll they suck. So, if you keep that in mind you can use a lot of their samples, except for snare. You might ask a really good drummer to come over and and hear the samples you've got. Just listen to snare samples. If they sound exactly the same when you're tapping on the keyboard very softly as they do when you strike the note as hard as you can, that's NOT a good sample.