Regarding the throttle accuracy in IL-2:
I'm not sure it jumps 4% when only going in one direction. I will make note of this within a few hours when I launch it again. Actually screw that, I'll go check right now
(edit confi.ini, change nohudlog=1 to 0. Test test. Check DxTweak2 for some more hardcore numbers).
Findings:
IL-2's full range of possible throttle positions is 0% to 120% - 120 or 121 steps.
The G940 potentiometers, when calibrated, has an output that allows for 130 to 150 different positions. There's a few millimeters between each 'position', and one must move extremely slow and careful to move it that little.
In game, the throttle setting (in numbers on the hud) as well as the physical throttle stick rendered in 3D, moved in 2% increments. every 5th or 8th step or something like that was a 1% increment.
Other axises, like the R1 and R2 on the throttle, has a wider number of possible positions. The Saitek Quadrant levers have 256 possible positions, and they track accurately in DxTweak2. Testing the most accurate of them, the Saitek Quadrant levers, showed that it would still frequently move in 2% steps, but that 1% steps occured more frequently.
Binding the stick itself, with sensitivity to the lowest (Meaning that big movement produces the least output), and moving it near the center, was the only way to get IL-2 to reliably move in 1% increments.
My conclusion is that there is something wonky about the way IL-2 reads the axis positions. IL-2 has a max of 121 positions and yet when it reads an axis with 256 positions (more than two per 'position') it does not translate accurately. The G940 throttle axises, with 130-150 positions, is more than one throttle position for each of the 121 possible throttle positions in IL-2.
----------
The reversing direction bug makes it jump 4% of the throttle in IL-2 for the G940. As I mentioned before, I saw that LogitechMark stated that the Logitech/Wingman team knew about this bug (firmware I think) and they are working to fix it in the next software update.
----------
For myself, 2% motion of IL-2 throttle is how it has always been like no matter the hardware (more or less). I turn off the unimmersive HUD log so I don't see those numbers when flying. Instead I look at the throttle position in the cockpit (it sometimes has markings indicating what use a position has), and I read the instruments. Speed, RPM, prop pitch settings.
As you mention getting double the size of the 'steps' in IL-2, I recommend calibrating the throttle unit, then checking out the readings from DxTweak2. Check the "RAW" readings in the lowest and the highest positions of the throttle, and then subtract the lower reading from the higher. The number you get is the number of positions each throttle axis can be registered as being in. Compare that to the numbers I posted above.