Gaussian copula has been employed to evaluate the outage performance of Fluid Antenna Systems (FAS), with the covariance matrix reflecting the dependence among multivariate normal random variables (RVs). While prior studies approximate this matrix using the channel coefficient correlation matrix from Jake's model, this work instead employs the channel envelope correlation matrix, motivated by the fact that the multivariate normal RVs are generated by transforming correlated channel envelopes. This raises an open question of whether using the coefficient- or envelope-level correlation matrix yields better accuracy in accessing FAS performance. Toward this end, this paper explores the benefits of using the envelope-level correlation matrix under fully correlated Nakagami-m fading, and develops a method for generating such fading channels for Monte Carlo simulations, which serve as a benchmark for validating the theoretical results. Simulation results confirm the effectiveness of the proposed channel modeling approach and demonstrate the superior accuracy of using the envelope-level correlation matrix, particularly in sparse port deployment and low-outage regime.