Email Post: Small population bias and sampling effects in stochastic mortality modelling