In principle, this in not surprising. Throw 100 normal-distributed values at the number line with mean 1 and std deviation 2 and some will end up below 0. The higher you make the mean, the fewer will end up negative and eventually it's unlikely (though still possible) one does.