I think this makes sense, but it's just unintuitive. Let's imagine you move the mouse 1 cm. During that time, the mouse takes a series of readings about it's location which then determine where the mouse should move on the screen.
If the mouse sensitivity is very high, it might take 1000 readings during that 1 cm. Each of those readings corresponds to a piece of data determining where the mouse moves.
Now suppose we set the sensitivity very low. Now it might only take 100 readings during the same 1 cm (because it's less sensitive to movement). But since these two actions correspond to the same movement of the mouse, the cursor will need to "jump" a little bit to fill in the "gaps" in data. That is, since your measurement is less fine, the movement on the screen is correspondingly jumpy, which might make it seem faster.
Personally, changing the setting between the maximum and minimum barely has any noticeable difference to me. But I imagine this has more to do with hardware (both the mouse and the display) than anything else.
I could be wrong, and maybe this is a hardware bug, but that's how I see it.