Accurate quantification of radiotracer uptake from lung PET/CT is challenging due to large variations in fractions of tissue, air, blood, and water. The air fraction correction (AFC) determines voxel-wise air fractions (AF) from the CT acquired for attenuation correction (AC) to correct for the variable air content. However, the resolution mismatch between PET and CT can cause artefacts in the AF-corrected image. In this work, we compare different methodologies for determining the optimal kernel to smooth the CT for AFC, when PET images were reconstructed with iterative reconstruction methods, and in the case where AF-corrected lungs exhibit uniform uptake. Noiseless simulations and non-TOF MLEM reconstructions were performed with STIR via SIRF using a digital test-object constructed from a CT scan of a patient with idiopathic pulmonary fibrosis. The optimal smoothing for AFC was determined via i) the point-source insertion-and-subtraction-method, h pts ; ii) artefact reduction in the AF-corrected volume-of-interest (VOI), h AFC ; iii) smoothing a ground-truth image to match the reconstructed image within the VOI, h PVC . Each of the three kernels were used to smooth the mu-map for AFC of the reconstructed emission data and the RMSE of each of the AF-corrected VOIs, with respect to the ground-truth, was assessed. Applicability of the preferred method was assessed with phantom acquisitions on a clinical scanner. Results demonstrated that, for iterative reconstruction methods, image resolution is dependent on iteration number and VOI density/uptake. Smoothing by h pts resulted in the least quantitatively accurate AF-corrected images at fewer than 40 iterations. For both h AFC and h PVC , RMSE in the AF-corrected IPF regions is fairly stable across iterations. h PVC has the potential to be utilised for determining the appropriate kernel for AFC on clinical scanners for application to patient PET/CT lung scans.