Axion-like particles (ALPs) are a well-motivated extension to the standard model of particle physics, and X-ray observations of cluster-hosted AGN currently place the most stringent constraints on the ALP coupling to electromagnetism, $ga\gamma$, for very light ALPs ($ma\lesssim 10-11$ eV). We revisit limits obtained by Reynolds et al. (2020) using Chandra X-ray grating spectroscopy of NGC 1275, the central AGN in the Perseus cluster, examining the impact of the X-ray spectral model and magnetic field model. We also present a new publicly available code, ALPro, which we use to solve the ALP propagation problem. We discuss evidence for turbulent magnetic fields in Perseus and show that it can be important to resolve the magnetic field structure on scales below the coherence length. We re-analyse the NGC 1275 X-ray spectra using an improved data reduction and baseline spectral model. We find the limits are insensitive to whether a partially covering absorber is used in the fits. At low $ma$ ($ma\lesssim 10-13$ eV), we find marginally weaker limits on $ga\gamma$ (by $0.1-0.3$ dex) with different magnetic field models, compared to Model B from Reynolds et al. (2020). A Gaussian random field (GRF) model designed to mimic $\sim 50$ kpc scale coherent structures also results in only slightly weaker limits. We conclude that the existing Model B limits are robust assuming that $\beta pl\approx 100$, and are insensitive to whether cell-based or GRF methods are used. However, astrophysical uncertainties regarding the strength and structure of cluster magnetic fields persist, motivating high sensitivity RM observations and tighter constraints on the radial profile of $\beta pl$.