Summary of our blinded B-mode investigation where we measure the COSEBIs E_n/B_n cosmic shear statistics, masking survey area where our astrometric metric, δr_exp, exceeds a series of threshold values (shown on the x-axis). The middle panel shows the p-value that quantifies the likelihood that the B-mode is consistent with zero. As we reduce the δr_exp threshold, the p-value increases above p ≤ 0.01 (red shaded region), and the masked survey passes the B-mode null test. As the mask reduces the total number of sources (lower panel), the blue squares show the p-values from a random mask analysis, demonstrating that the improved B-mode is not driven by a reduction in area. In the upper panel we find that the astrometric masking has a benign impact on our Σ₈ cosmological constraints. The green data point and shaded regions shows the results from our fiducial analysis, demonstrating that failing data-vectors are consistent with our fiducial, all having less than 0.4σ changes in Σ₈.