Same as Fig. C.1, but for HD 167946 (observation X0460). Wavelengths below 1.03 μm are affected by large flux calibration errors. Between 1.03 and 1.1 μm the situation is unclear: while flux calibration errors may be responsible for some of the discrepancy seen for this spectrum, we note that a similar discrepancy is seen in most hot stars (out of 21 XSL spectra between 9000 and 13 000 K, only two have fluxes above those of the model in that region), while the discrepancy tends to be the other way round in cool stars (empirical fluxes higher than the model fluxes).