The semiconducting and self-assembling properties of columnar discotic liquid crystals have stimulated intense research toward their application in organic solar cells, although with a rather disappointing outcome to date in terms of efficiencies. These failures call for a rational strategy to choose those molecular design features (e.g., lattice parameter, length and nature of peripheral chains) that could optimize solar cell performance. With this purpose, in this work we address for the first time the construction of a realistic planar heterojunction between a columnar donor and acceptor as well as a quantitative measurement of charge separation and recombination rates using state of the art computational techniques. In particular, choosing as a case study the interface between a perylene donor and a benzoperylene diimide acceptor, we attempt to answer the largely overlooked question of whether having well-matching donor and acceptor columns at the interface is really beneficial for optimal charge separation. Surprisingly, it turns out that achieving a system with contiguous columns is detrimental to the solar cell efficiency and that engineering the mismatch is the key to optimal performance.