Left: cumulative distribution functions of fixing G_N (red line) or imposing empirical lognormal priors (blue line) and flat priors (black line) on G_N. The three cases show indistinguishable fit qualities. Right: histograms of the maximum-likelihood G_N. The dark and light blue histograms correspond to lognormal and flat priors on G_N, respectively. The inset (right panel) shows a zoom-in on the distribution for the Gaussian prior, switched to linear scale for a better view. The actual value of G_N is indicated by red dashed lines. The Gaussian prior returns a tight distribution of G_N without decreasing the fit quality; this confirms the constancy of G_N as expected. The same argument has been applied for the constancy of g_† in Li et al. (2018).