[en] Many quantitative studies of relativistic cosmic ray propagation exist in which ''standard'' values for the input quantities are adopted in an uncritical manner. In contrast, the major emphasis of this study is on developing the proper set of formulae and error estimates for each of the atomic and nuclear processes that govern the composition of the cosmic rays between lithium and nickel. In particular, it is shown that errors of approximately a factor of two exist in the standard (Bohr) cross sections for stripping, that the correction function from high energy photoionization needs to be introduced into the standard cross section for radiative attachment, and that because the half-life of a fast nucleus with at most one K-shell electron can differ from the half-life of a neutral atom, several laboratory-based values need correction. The framework used to assemble and correct these quantities is a matrix formalism for the leaky box model similar to that used by Cowsik and Wilson in their ''nested leaky box'' model. It is shown that once the assumption of species-independent leakage is introduced, the matrix formalism becomes virtually identical with the standard exponential path length formalism. 87 references