Why throw out the low and the high? Aren't they relevant to the average? Throw out results if you believe you skewed them in your testing; the rest is relevant to estimations of variance in the numbers.
In fact, the std deviation from the mean (and %rel. error) turns out to be one of the things most useful to show when collecting data. Take those and make error bars on your points for the results. Make x and y error bars to account for flow rate error too and you'll start seeing why perpetual equipment upgrades are a reality for some.
|