Effect of different b(m) prior distributions on the reconstructed r(k) for CFHTLenS-like noisy data (the orange ‘new prior’ and the blue ‘old prior’) in comparison to the green SH18-like data with high signal-to-noise ratio. The posterior constraints indicate 68% and 95% credibility regions. The dashed line is the true r(k) of the H15 semi-analytic galaxies used in this test (SM4 high-z).