A self-consistent Schrodinger-Poisson solver is used to improve upon a recent evaluation of the attainable DC performance of coaxial carbon nanotube field-effect transistors. The earlier evaluation, which was based oil the predictions of a compact model, is shown to be optimistic because of the model's inadequate treatment of quantum mechanical reflection of thermionically injected carriers. This deficiency of the compact model is remedied, to a large extent, by incorporating a new, short expression for quantum mechanical reflection under phase-incoherent conditions.