Your fan experiments are very curious. I have to admit I pretty much wrote off the 120x38 fan after listening to it separately and then seeing the results of the testing. But other questions arise concerning your build:
1) Why would you put the HDDs in the upper cage and not the lower? The lower HDD cage is better for cooling & also for noise.
2) What's the rest of the system?
Hard drives are in the upper cage because they are PATA and the ribbons don't reach the lower cage. I finally purchase a motherboard where the ATA headers aren't on the bottom corner of the board (which made cable routing a pain in my old case), and then I get a case where that would be the perfect place for 'em.
The system contains:
Sempron 2600+ @ 2.1GHz, cooled by Thermalright XP-90 with Panaflo L1A 80mm @ ~7V in "suck" orientation;
6600GT AGP video card with stock fan replaced by Panaflo L1A 80mm @ ~9V;
Seasonic Super Silencer 300W, Rev. A3;
1604N and 1614N Samsung Spinpoints;
The components, ranked from noisiest to quietest, are:
hard drives (even one by itself might be the loudest component);
9V Panaflo on video card;
5V-high 120x38mm case fan;
7V Panaflo on CPU;
either 5V-medium 120x25mm case fan.
Update: I was mistaken; the ribbons DO reach the hard drives in the lower cage. But drive temperatures went up 10C compared to when they were fan-cooled in the upper cage, even though I (mostly) taped the vent holes around the PSU. The hard drives weren't spaced apart, but I doubt that moving the drives to opposite sides of the cage would make a big difference. I couldn't squeeze the 120x25mm fan into the space between the hard drives and power supply without undoing half an hour's worth of careful PSU wire routing, so I gave up and put the drives back in the upper cage.
I'm not sure how much difference the choice of drive cages affects noise. Devon says at the bottom of Page 8 that both noise and cooling are virtually identical for both cages, although the bottom cage cools better when the PSU vents are taped.