The actual optimal solution is 100%MM or the Gear Ratio method. But because of distortion, it doesn't feel correct at the crosshair.
Muscle memory is both distance and speed. But really your muscle memory is about moving your mouse and expecting to move to a certain point on the monitor. But because of distortion, you are never going to maintain muscle memory across all points on the monitor when changing your FOV. The best thing you can do is choose the method that you think will benefit you the most, and do some experimenting.
0% is matched at the crosshair so it's good for tracking and making adjustments, but it isn't as good with flicks closer to the edge of the monitor.
100% is matched at the edge of the monitor so flicks are easier closer to the edge. And as Drimzi mentions below, the sensitivity is more uniform across the entire FOV range.
Every other monitor match percentage is a combination of 0% and 100%.
Viewspeed hangs around 75%MM. I don't really know why you say it's the optimal solution... FYI, the description of "Viewspeed" actually applies to the Gear Ratio Method, which is 100%MM. And I agree, 100%MM makes the most sense and would be the optimal, perfect solution if it weren't for distortion.