Performance of the inversion method as a function of the true polarization signal for different cloud distances, probed by the values of f_bg, shown in the legend and given in per cent. The method is applied to simulated data from the 1° aperture star sample (345 stars). Top: relative difference between the cloud parallax at maximum log-likelihood with the parallax of the closest star to the input cloud . Middle: L₂, distance between the true and estimated mean polarization vector. Bottom: ξ, the ratio between ideal and estimated posterior size on clouds’ Stokes parameters. P_max varies in multiples of 0.05%, and A_turb = 0.2 and 10 simulated samples are obtained by varying . For the same f_bg, the solid lines connect the median of the 10 reconstructions on simulated samples with the same P_max. Filled (empty) symbols correspond to fits that pass (do not pass) the criterion explained in the text.