peeush,
I don't see the flight test relevance to your question.
It seems like you're trying to use a normal distribution on measurement errors, by assuming that the maximum range of measurement errors is equivalent to three standard deviations, and from there deducing that half of your measurements have a measurement error of one third of the maximum.
For that to work you need to identify all sources of measurement error and confirm that each is error is idendependant of all others. Once you've done that, you can claim a probability that 50% of the measurements are within n/3, but you cannot state that it is so.
On average 50% of coin flips will turn up heads, that doesn't mean you won't ever see 10 heads in 10 flips.
Matthew.