Lunar occultation data often are reduced by means of a multiparameter least-squares fitting routine. For binary stars the minimum number of free parameters is five: the time of geometric occultation for each star, the intensity of each component, and the background intensity. From these values the magnitude difference between components can be calculated as well as the angular separation of the binary in the direction perpendicular to the lunar limb (i.e. the lunar position angle). If the observed rate of motion of the lunar limb can be determined as an additional free parameter, then the difference between observed and predicted values can be interpreted in terms of a local lunar slope. Therefore, in principal the observed rate of motion can be used to correct the vector separation — both angular separation and direction — of the binary. In this paper four occultation binaries observed from KPNO are examined to determine whether binary-star astrometry can be improved using these corrections.