We present a master equation study of organic nitrate formation from the peroxy radical (RO2) + NO reaction. The mechanism is constrained by both quantum chemical calculations of the potential energy surface and existing yield data. This mechanism displays heretofore unrecognized features of the system, including distinct conformers of a critical peroxynitrite (ROONO) intermediate that do not interconvert and a dual falloff behavior driven by collisional stabilization in multiple wells. These features have significant implications for atmospheric chemistry; in particular, only a fraction of the ROONO intermediates may easily isomerize to nitrates, resulting in a limit to total nitrate production. Existing mechanisms, extrapolated to low temperature and high pressure, produce nitrate almost exclusively. As a consequence, hydrocarbon oxidation sequences based on these mechanisms do not propagate radical chemistry, which is inconsistent with available experimental data. To reproduce observed nitrate yields, we model a transition state from the ROONO intermediate to RONO2 that differs considerably from the few found in computational studies. Specifically, the data require that this transition state energy lie well below the energy of separated radical products (RO + NO2), while computational studies find the transition state at higher energies. A second feature of yield data is difficult to model; to enable collisional stabilization of C-5 systems, as observed, we reduce the unimolecular decomposition rate constants from the ROONO intermediate by a factor that is at the far end of the plausible range. However, with these experimental constraints in place, the model successfully reproduces multiple features of existing data quantitatively, including both high-and low-pressure asymptotes to nitrate production as well as the observed shifting of pressure falloff curves with carbon number. Consequently, we present a new parametrization of nitrate yields, providing interpolation equivalent to existing parametrizations but dramatically improved extrapolation behavior.