Lin / Log pot differences could cause the observed sensitivity issues, but if I had the problem, I think I'd consider PWM drive for the motor. The mass of the magnets would smooth out the rotation just by inertia, and I think you'd have really good control.
You could use a micro to do it or if you wanted to stay in the analog realm you could use a 555 to drive a pass transistor or even the enable on a regulator.
If you REALLY wanted to do it right you could use a hall-effect device as part of a feedback loop so that you don't lose your magnetic coupling to the stir bar.
I can whip up a schematic if someone wants to prototype it.


<see, it's getting better.


