A really small datasets shouldn’t be used for statistics. But is it small, considering there are millions of flights registered?
More concretely: if you flip an unfair coin 1 million times and get a single heads, you know that the odds of getting that heads are extremely low, but you can't yet say what the odds are. It's possible that the odds are one in a million, but it's also possible that you got very lucky or very unlucky relative to the actual odds. You have to have a lot of data in both buckets before you can distinguish.
All this dataset provides is that with small enough samples it is too noisy to draw reliable conclusions.