You made me realise that the graph y-axis name is missleading. It's actually the output force. I ran the linear force test while holding the wheel tight. Even if the wheel didn't move at all, i could feel that the applied lateral torque was going stronger and stronger.
So even if the data is reported as a rotation, it's actually a lateral torque ramp up.
I did a small graph that would matche yours.
In that graph, i recalculated the output forces as percentage instead of raw values. It give a slope that is fairly similar to the first one i posted. With that said, the Knee is still at 0.38. That's a given. Now, for the Reduction calculation. Before the math, with your example, we can see applying some scoop will bend a linear response downward. If we apply some reduction on a upward bend, it should straighten that line.
Now, for the maths, since i converted the data, here it goes.
With Deadzone removal taken into account. I get a Scoop Reduction of 0,25
Without Deadzone removal, i get a Scoop Reduction of 0,13
I'll try to post an updated sheet later on. Or if someone want to update it!
What do you think?