This paper presents new results and estimation bias analyses for a Bayesian mark-recapture model, applied to the South Georgia toothfish tagging data. The updated model incorporates the data from all the analysed cohorts in the estimation routine, and we examine the potential estimation bias, using simulated data. The results confirm earlier findings, with lower estimates of natural and fishing mortality, for the given age range, than are assumed and predicted by the current assessment, respectively. Also, we find some apparent estimation bias, which is similar, but lower than is seen in similar mark-recapture models of this type. Given the increasing number of releases and recaptures in all the current tagging programs, we suggest that this type of modelling approach can serve as both a useful tool for estimating key parameters like natural mortality, as well as a method for comparing parameter estimates of the exploitation rate predicted by the full assessment models, currently being implemented.