Most of the neutronic core calculation schemes used by industrialists for nuclear reactor studies are based on a two-step deterministic scheme: A two-dimensional transport calculation at the assembly level produces homogenized and condensed nuclear data used by a full three-dimensional core solver under the diffusion approximation. The validity domain of such schemes is driven by the hypotheses of the diffusion approximation, implying wide meshes (10 to 20 cm), and is hence limited to studies that do not require a thinner description of neutron behavior. Consequently, local phenomena and heterogeneous interfaces are not yet fully validated with industrial two-level calculation schemes. For instance, the axial core-reflector interface, which is characterized by extra thermalization of neutrons leading to a local (≈2-cm height) increase of the neutron flux at the axial edges of fuel pins, is specifically challenging for deterministic methods. To cope with this issue, specific safety studies are performed with reference Monte Carlo simulations. This paper shows that enhancing the equivalence method enables flux discrepancies to be reduced from 12% to 6% for mixed oxide fuels and from 9.3% to <1% for uranium oxide fuel (impacting the power discrepancy from 5% to 3% and from 0.3% to almost 0.0%, respectively) between Monte Carlo and deterministic simulations (SCIENCE V2). The improved equivalence method uses dedicated discontinuity factors and constants, according to an optimized mesh composed by a mesh for each medium and a refined mesh in the fuel region.