Gas diffusion (i.e., Knudsen diffusion) through nanopores is an important research due to its significance in the areas of catalysis, gas separation, transport through carbon nanotubes, graphene slits as well as tight shale rocks. The mass flow rate has been widely predicted by the Knudsen diffusion model proposed in 1909, which is based on the Fick's first law using the self-diffusion coefficient. Gas transport enhancement through nanotubes, relative to the prediction by the prevailing century-old Knudsen diffusion model, is commonly reported in Science, Nature and PRL. This enhancement is usually attributed to the partly specular molecular reflections at the smooth nanotube surface (characterized by the surface accommodation coefficient), which break the model assumption of completely diffusive reflections (fully accommodated). However, an oversighted cause of the discrepancy between the measurement and theory that we found is that even for the gas transport with completely diffusive reflections, the Knudsen diffusion model based on the Fick’s first law is accurate only for very long nanotubes. Additionally, for smooth nanotubes with partly specular reflections, the Knudsen diffusion model is also invalid even if the self-diffusion coefficient is corrected to account for the atomic-scale surface smoothness. On the other hand, the Knudsen diffusion model might be used for interpretations instead of predictions, and then the transport diffusion coefficient inferred from the measured mass flow rate could be completely different from the self-diffusion coefficient. All those discrepancy and confusion stem from the implementation of the Fick’s first law and can be avoided by using the molecular transmission probability obtained by Berman using the kinetic theory to quantify the flow rate of the Knudsen diffusion process. This work provides the correction to the Knudsen diffusion model for accurate predictions of gas diffusion through nanotubes and better interpretations of experimental measurements.