There is a large literature that reports oil-water partition coefficients of small molecule compounds. A goal of much of this work is to obtain hydrophobicity parameters for models of folding, docking, binding, conformational changes, partitioning, and flow, mainly of biomolecules and drugs. The quantity obtained from experiments is the difference between a solute's oil- and water-phase chemical potentials Delta mu = mu(wat) - but the quantity needed For models is a contact free energy per unit area. Usually, solvation modelers find this quantity in two steps: (i) subtract off some assumed "irrelevant" solute entropies from the experimental Delta mu's and then (ii) divide the result by the solute area. Here we show that such a procedure can lead to Large errors. We use a two-dimensional lattice model of hydrocarbon liquids in NVT Monte Carlo simulations, where the underlying contact energy is rigorously known in advance, to test strategies for converting partitioning data to contact energies. We find that rather than using mu(wat) - mu(oil) and subtracting assumed entropies, it is better to use mu(wat) - h(oil), where h(oil) is the solute's enthalpy of solvation in the oil phase. We apply this strategy to experimental thermodynamic data from liquid alkanes and propose that the best estimate for the microscopic free energy of transfer of a short-chain hydrocarbon from oil to water is around 40 cal/(mol Angstrom(2)).