In a "regular" fusee arrangement -- with the fusee on one arbor and the mainspring on another arbor -- that 10% power decrease per turn power would have to be spread over a variable number of turns of the fusee, depending on its radius at that point. At its small end, the fusee has to make more turns to wind a single turn off the mainspring than it does at its large end, so the curve on the small end must be flatter. Which is why the curve on a "regular" fusee is a complex curve, not linear.