I recorded all of my runs and then using some "crude" statistical analysis you discard the one or two highest and lowest readings, average the rest. I'd try to get at least 8 very similar runs to establish your baseline.

I wouldn't worry too much about HP and Torque I wasn't able to get consistent results from my G-Tech Pro. I focused more on 1/4 mile time and speed. Unless you can get good, consistent results with the power readings.