Aims. We investigate the performance of a deconvolution map-making algorithm for an experiment with a circular scanning strategy, specifically in this case for the analysis of Planck data, and we quantify the effects of making maps using simplified approximations to the true beams. Methods. We present an implementation of a map-making algorithm which allows the combined treatment of temperature and polarisation data, and removal of instrumental effects, such as detector time constants and finite sampling intervals, as well as the deconvolution of arbitrarily complex beams from the maps. This method may be applied to any experiment with a circular scanningstrategy.Results. Low-resolution experiments were used to demonstrate the ability of this method to remove the effects of arbitrary beams from the maps and to demonstrate the effects on the maps of ignoring beam asymmetries. Additionally, results are presented of an analysis of a realistic full-scale simulated data-set for the Planck LFI 30 GHz channel. Conclusions. Our method successfully removes the effects of the beams from the maps, and although it is computationally expensive, the analysis of the Planck LFI data should be feasible with this approach.