So the quoted uncertainty is usually a lower bound on the true uncertainty, and it is likely a few times larger in reality (my very unscientific "look briefly at data and methods" made me think claiming anything below 5% uncertainty would be embarrassing regardless of sample size, but YMMV).