Slim
The posters who have pointed out the necessity of relating small probability and severity of consequence have it right.
This is the key. I live next door to an 18,000 hour plus ATPL who has never had an engine failure.
Bluskis has had some hairy moments.
You are looking at numbers across a huge sample and all you are seeing is the result of a normal distribution - 98% of the variance will be contained in 2 standard deviations.
If the other 2% doesn't concern you, then you have established a persoanl risk tolerance that is usable.
But it doesn't alter the probability/severity relationship