Well, not so normal...
I was able to try 2010 firmware and 2009 firmware on a same NGC4 PCM and vehicle. 2010 firmware does it, 2009 does not. I struggled a long time trying this and that to figure out why, and what are the differences. And they are various, but nothing helped. 2009 firmware settles the 2C CVI to around 61, 4C to around 38, does not do 4-5 hesitation, but instead does 1-2 and 4-5 bumpalong. 2010 firmware goes to 4C CVI of 25-28, and 2C CVI of 47-48, does not do 1-2 and 4-5 bumpalong (or does it much less), but does 5-4 hesitation. In the 2010 code, there are many new things - for example time-to-slip is learned alogn with the CVIs, and much more.
The thing is, my packs are upgraded 2C (Alto red eagle + kolene and additional steel and friction), upgraded 4C (Raybestos GPZ, also with additional steels and frictions), upgraded UD (from 68RFE 1-side discs, raybestor GPZ), and OD (Again 68RFE 1-side discs, Raybestos GPZ).
While fighting 1-2 (mostly) and 4-5 bump-along, since 1-2 shift is purely 2C pack engaging without outgoing pack (LR pack is disengaged long before shift and only 1-way roller clutch is active), it was clear that 1-2 bumpalog was purely because something with 2C clutch engagement was not right. It could be either too slow or too fast engagement, and knowing that the pack is more agressive than standard, after playing with pressures and other things, which didn't do anything to solve the problem, I changed the upshift acceleration limits. Why? In WOT condition, bump-along was gone (the WOT limit is -8500 from factory on 2009 FW), in light throttle condition it was masked by the torque management, and only happened in mid to high throttle. I changed 1-2 A limits to -400, -2800 and -4000. It improved things alot. Then changed them to -400, -3400, -4500. I will able to say in a few days, but today I was not able to make it perform the bumpalong anymore, no matter what I did. And it was easy to make it do it before.
Next, I thought the 5-4 downshift must be something similar, but there's nothing mapped anywhere for downshift acceleration that I know of, and I had to analyze the decompiled code to find it. When I did - BINGO! There was the difference between 2009 firmware that does not have this behaviour, and 2010 that does. They reduced 5-4 and 4-3 desired acceleration considerably. I don't know the reason, maybe it works fine with the OEM steels and frictions, but it definitely didn't work with upgraded 4C pack from Raybestos. Maybe they tried to be more gentle on 4C and OD packs. I don't know.
Take a look at pictures. They say much.

I changed 4-3 like in 2009, and 5-4 to 3000 (2500 in 2009 FW) and 1200 (1000 in 2009FW). I was not able to make it do the 5-4 hesitation anymore. It makes the shift right when it decides. First it unlocks the TCC, then half a second later it performs a fast and firm 5-4 shift.
I don't have any problems with 4-3 shifts, so maybe 1-side pack from 68RFE performs closer to the OEM. The CVI of UD and OD are the same on 2009 and 2010 FW.
So, problems solved. It seems that when non-standard packs are used, non-standard tune should be used along with them.
Maybe playing with things like initial DC and various slopes/gains could do the same, but I want more firm shifts anyway, so acceleration limits/desired acceleration are OK for me. It shifts great with torque management, and in WOT. Maybe I will try to turn Tq management off to see what happens - it maked the 1-2 bump-along much worse in the past. I am OK with behaviour for the moment, so I don't know if/when I will try it.