Jamie:
There is a discrepancy because the sampling is time fixed and the input is time variable, and the window chosen allows for limited peak measurement.
When the system is loaded every thing is OK because You are stabilizing the wind mill for one ( You are placing a semi brake to it, allowing more uniform output power).
The response of the Hall Effect if too fast for You, slow it down by placing a low past filter in the output, for a wind mill 2 to 5 Hertz low pass is OK, all depending on the response time of the wind mill.
At average wind velocity, what is the generated frequency of the wind mill ?.
Divide it by 3 or 4 and that could be the Current detector low pass filter -- I am guessing here, because I do not know your over all system parameters.
Nando