It does indeed have two peaks: https://en.m.wikipedia.org/wiki/Inverse_distribution#Recipro...

As the mean of this distribution does not exist (it's "infinite") the Monte Carlo estimates aren't reliable