Notes. Stage III survey forecasts were conducted with various sets of cosmological probes, redshift biases, and redshift recalibration nuisance models (left-most columns; see also Table 2). The remaining columns give: the posterior mean parameter bias (in units of σ, see the end of Sect. 4) on S₈, or on the S₈ − Ω_m plane (Eq. (4.2)), relative to the idealized case ($\mathrm{probes}:n_z^{\mathrm{true}}:\mathrm{do}\mathrm{nothing}$; boldface rows); the inverse error 1/σ₆₈ on S₈, or the S₈ − Ω_m FoM; the best-fit χ_d² at the MAP parameter set (Eq. (2.23)); and the S/N of the data-vector (hence the numbers for different nuisance models are the same when sharing the same data-vector). Numbers for forecasts conducted with exactly zero redshift bias ($n_z^{\mathrm{true}}$) are given in boldface. As expected, we see a perfect recovery of the fiducial parameters by the MAP when the bias is exactly zero, signified by zero χ² (some biased shear-only cases show zero χ² due to rounding).