I think the artifacts could be due to the way you handled the data. I assume the dyno data is put in every x rpm. In effect this discrete data is a step function every x rpm. When the derivative is taken that becomes a delta function. Which would explain the regularity of those spikes.