That's assuming that the distribution is purely gaussian and nothing weird happens at the tails.

I agree 276 is unlikely (and how would you even test/norm such a thing?)