Generalized SCIDAR (GS) measurements, which were taken at the Paranal Observatory in 2007 November/December in the context of a site qualification for the future European Extremely Large Telescope, are recalibrated to overcome the bias induced on the ? profiles by an incorrect normalization of the autocorrelation of the scintillation maps, which has recently been identified in the GS technique. A complete analysis of the GS-corrected measurements and the corrected errors is performed statistically as well as on individual nights, and for each time period during the nights. The relative errors of the ? profiles can reach up to 60 per cent in some narrow temporal windows and some vertical slabs, with the total seeing up to 12 per cent and the total integrated turbulence J up to 21 per cent. However, the statistical analysis shows that the absolute error of the median values of the total seeing is 0.06 arcsec (relative error 5.6 per cent); for the boundary seeing it is 0.05 arcsec (relative error 5.6 per cent) and for the seeing in the free atmosphere it is 0.04 arcsec (relative error 9 per cent). We find that, in spite of the fact that the relative error increases with the height, the boundary and the free atmosphere seeing contribute in an equivalent way to the error on the total seeing in absolute terms. Moreover, we find that there are no correlations between the relative errors and the value of the corresponding seeing. The absolute error of the median value of the isoplanatic angle is 0.13 arcsec (relative error 6.9 per cent).