In this paper, we address the problem of planning the crude distillation unit charging process with oil blend. It is well known that blending and splitting operations can lead together to both non-linearities and concavities in mathematical programming models. As result, many proposed models for this problem use simplifying assumptions to keep the formulation computationally tractable. However, we show the existence of splitting operations that can lead to inconsistencies in the solutions obtained by the previous MILP models from the literature. Then, we propose a way to address this issue through an aggregated inventory capacity combined with a disaggregation algorithm. Furthermore, we develop a mathematical reformulation that improves the solving efficiency of the method. Then, we report experiments that show that the reformulated MILP model presents significant gains concerning linear relaxation gaps and run times, and the disaggregation algorithm leads to feasible solutions for all the tested instances. (C) 2015 Elsevier Ltd. All rights reserved.