Motion compensation (MoCo) is pivotal in the processing of airborne synthetic aperture radar (SAR) data. Motion errors are space variant according to the data acquisition geometry, which can be split into range-variant and azimuth-variant components. In the processing of wide-beam SAR data, the azimuth variance must be considered. Several approaches have been proposed, among which the subaperture algorithm based on frequency division (FD) is a good choice if motion errors involve high-frequency components. However, there are two drawbacks to using this algorithm in conditions where the magnitude of motion errors is large, i.e., the invalidation of the time-frequency relation in an FD with too many subapertures and paired echoes caused by periodic discontinuities in the azimuth phase history. Theoretical analysis and simulations are presented to demonstrate these two drawbacks. Improvements are then made on the traditional algorithm: implementing the FD after the subaperture MoCo instead of before it and adopting a nonuniform FD scheme rather than a uniform one. Finally, the validity of the improved algorithm is further demonstrated by experimental results with real data.