A number of models have been developed to predict rock permeability using the mercury injection capillary pressure (MICP) data. Liu et al. (2016) argued that the classical Swanson (1981) and the capillary parachor models are not precise, as the models did not consider the effect of the porosity. However, a few issues may exist in the Liu et al. (2016) model. First, 30 core samples were used to calibrate the model parameters, but the same core samples were reused for the model validation. Second, the model is dominated by the contribution of porosity rather than the Swanson parameter and the capillary parachor. Third, the authors processed all 30 core samples together despite the fact that they were from different formations. The disorder of the core samples makes the classical Swanson and the capillary parachor models are no longer applicable. We find that the core samples can be divided into two different clusters according to a critical porosity of 28.0% from their data, and then we can directly use the classical Swanson and capillary parachor models to estimate the permeability without considering the porosity. Both the Swanson and the capillary parachor models work well in the conventional reservoirs, but they are not appropriate for the unconventional reservoirs such as the tight sandstone reservoirs because it is very difficult to obtain the Swanson parameter and capillary parachor. At the same time, we find the average pore throat radius (Rm) is strongly related to permeability due to the similar distributions of the pore throat radii of those 30 core samples. Therefore, we use the Rm to establish an alternative permeability prediction model. Our findings are significant in establishing a reliable permeability prediction model using the MICP data.