Median and mean are both averages, and there are other types of averages. Average by itself doesn’t imply mean. And of course it’s median - how can you have a mean of something discrete like group of people?
The arithmetic mean (I.e. the average) is the 50th-percentile (I.e. the median) in a normal distribution. The question absolutely stands - unless pilot ability is not a normal-distribution.
In anything with a lower bound (passing to become a pilot) the distribution can be heavily skewed (ie cutting off the bottom 20% of a normal distribution)
Which means that more than half would be below average!
Since there is a long training/weeding-out process I would expect the distribution would be truncated normal, with some cutoff to graduate plane school. If this bar is low then the distribution will still be roughly normal, but if it's near or about the mean (does anyone have the fail rates?) then the distribution can look a lot different, e.g. one-sided normal, or power-law.
Why would a competitive, high skilled role with aggressive disqualification requirements be anywhere close to a normal-distribution?