Top row: models of the mock data shown in Fig. 6. Middle row: reconstructed perturbations. Bottom: model residuals in normalized units, i.e., divided by the diagonal, σ_d, of the data covariance matrix from Eq. (14). We note in particular that within the area of the most significant lensed flux, as indicated by the mask (solid colour), the residual are identical to noise. The reduced chi-squared of the models is also given in the panels of the bottom row.