The study of natural CO2 storage reservoirs has demonstrated that carbon mineral trapping is the safest mechanism for subsurface CO2 storage. The slow rate of fluid-rock reactions makes it imperative to use reactive transport simulations as a tool for estimating reservoir scale carbon mineral trapping capacities. However, there is often a disparity in the estimations of carbon mineral trapping capacities determined from reactive transport simulations and recorded in natural CO2 storage reservoirs. This might be due to several factors such as the uncertainty around simulations parameters such as mineral reactive surface areas as well as the neglect of sub-core scale lithological heterogeneity in reservoir scale simulations. This study points to the importance of the latter to reconcile the mismatch. Multiphase reactive transport simulations were run on a cm-scale resolution geological model (CRM) of the Paaratte Formation, Otway Basin (Australia), where mm- to cm-scale lithological heterogeneity is incorporated via upscaled rock parameters. Simulations were also run on a m-scale resolution model (MRM) of the same transect to quantify the difference in the predicted secondary mineral composition due to the incorporation of sub-core scale heterogeneity in reservoir scale simulations. The simulation outcomes for both the models were compared to the data from the Pretty Hill Formation, a natural CO2 storage analogue within the Otway Basin with similar primary mineral composition. The results from the CRM compare well with the secondary carbonate mineralogy of the Pretty Hill Formation. In contrast, the MRM strongly underestimates carbonate mineral formation. The results highlight that the incorporation of mm- to cm-scale lithological interfaces in reservoir scale reactive transport simulations can increase carbon mineralization estimates in siliciclastic formations by 3–6 orders of magnitude as these interfaces are sites for mineral precipitation.