In the diffusion region of magnetotail reconnection, particle distributions are highly structured, exhibiting triangular shapes and multiple striations that deviate dramatically from the Maxwellian distribution. Fully kinetic simulations have been demonstrated to be capable of producing the essential structures of the observed distribution functions, yet are computationally not feasible for 3D global simulations. The fluid models used for large-scale simulations, on the other hand, do not have the kinetic physics necessary for describing reconnection accurately. Our study aims to bridge fully kinetic and fluid simulations by quantifying the information required to capture the non-Maxwellian features in the distributions underlying the closures used in the fluid code. We compare the results of fully kinetic simulations to observed electron velocity distributions in a magnetotail reconnection diffusion region, and use the maximum entropy model to reconstruct electron and ion distributions using various numbers of moments obtained from the simulation. Our results indicate that using only local moments, the maximum entropy model can reproduce many of the features of the distributions: (1) the anisotropic electron distributions inside the ion diffusion region but outside the current-sheet can be modelled with 10-14 moments, (2) the electron-outflow distribution with a tilted triangular structure is reproduced with 21-35 moments and (3) counterstreaming distributions can be captured with the 35-moment model when the separation in velocity space between the populations is large.