In our paper, it was shown that dense red blood cells, that are supposed to be less deformable than the others, have a smaller migration velocity than less dense cells. However, due to a mismatch in datasets, an incorrect set of data was plotted in Fig.5 for the lift of the least dense cells-the mistakenly plotted data corresponded to an average behaviour of the whole suspension. As a matter of fact, the correct data shows an even stronger migration than reported initially, which actually strengthens the conclusions of the paper. From the fit of the data, the lift parameter $/xi$ is found to be 0.018 for the least dense cells (while we reported 0.014 in the original paper), to be compared with 0.0065 for the densest ones. The authors would like to apologise for any inconvenience caused. New Figure 5: similar to original Fig 5 with correct data for the least dense cells (red dots).