You had mentioned 1e6 RPM in your original message, so that would obviously be an erroneous reading.
So eliminate any values that are too large, to start with.
Then following that with your averaging and percentage change seems like a way to take care of the false, but otherwise plausible readings.