Sediment transport equations used in wind erosion and dust emission models generally incorporate a threshold for particle motion (u*t) with a correction function to account for roughness-induced momentum reduction and aerodynamic sheltering. The prevailing approach is to adjust u*t by the drag partition R, estimated as the ratio of the bare soil threshold (u*ts) to that of the surface in the presence of roughness elements (u*tr). Here, we show that application of R to adjust only the entrainment threshold (u*t = u*ts/R) is physically inconsistent with the effect of roughness on the momentum partition as represented in models and produces overestimates of the sediment flux density (Q). Equations for Q typically include a friction velocity scaling term (u*n). As Q scales with friction velocity at the soil surface (us*), rather than total friction velocity (u*) acting over the roughness layer, u*n must be also adjusted for roughness effects. Modelling aeolian transport as a function of us* represents a different way of thinking about the application of some drag partition schemes but is consistent with understanding of aeolian transport physics. We further note that the practice of reducing Q by the vegetation cover fraction to account for the physically-protected surface area constitutes double accounting of the surface protection when R is represented through the basal-to-frontal area ratio of roughness elements (σ) and roughness density (λ). If the drag partition is implemented fully, additional adjustment for surface protection is unnecessary to produce more accurate aeolian transport estimates. These findings apply equally to models of the vertical dust flux.