Comparison of the EE, BB, and EB angular power spectra for the emulated maps (red) versus the reference dataset (blue), shown for the N_input = 10 (left) and N_input = 100 (right) cases. In each row, the top panel displays the average power spectra with 1 σ shading, while the bottom panel shows the residual, ${\mathrm{\Delta }}_{\ell }=(⟨C_{\ell }^{\mathrm{e}\mathrm{m}\mathrm{u}}⟩-⟨C_{\ell }^{\mathrm{r}\mathrm{e}\mathrm{f}}⟩)/{\sigma }_{\ell }^{\mathrm{r}\mathrm{e}\mathrm{f}}$ (solid red) along with ${\mathrm{\Delta }}_{\ell }^{\pm }=(⟨C_{\ell }^{\mathrm{e}\mathrm{m}\mathrm{u}}⟩-⟨C_{\ell }^{\mathrm{r}\mathrm{e}\mathrm{f}}⟩\pm {\sigma }_{\ell }^{\mathrm{e}\mathrm{m}\mathrm{u}})/{\sigma }_{\ell }^{\mathrm{r}\mathrm{e}\mathrm{f}}$ (dotted red), where ${\sigma }_{\ell }^{\mathrm{r}\mathrm{e}\mathrm{f}}$ is the standard deviation from the reference set and ${\sigma }_{\ell }^{\text{emu}}$ is the emulated dataset's one. In both the 10-map and 100-map cases, the mean spectra and variance per multipole closely match those of the reference set, and the residuals lie well within the 1σ band. The bottom panels also show in dotted gray the standard error on the mean for a sample of size N_input, that is $\pm {\sigma }_{\text{ref}}/\sqrt{N_{\text{input}}}$, with N_input = 10(100) in the left (right) plots. Notably, the scatter of the mean residual is roughly within this band in both cases, indicating the success of the augmentation process. See Section 5.3 for further details.