Deriving oxygen abundances from the electron temperature (hereafter the$T_e$-method) is the gold-standard for extragalactic metallicity studies.However, unresolved temperature fluctuations in HII regions can biasmetallicity estimates low, with a magnitude that depends on the underlying andtypically unknown temperature distribution. Using a toy model, we confirm thatcomputing $T_e$-based metallicities using the temperature derived from the [OIII] $\lambda$4363/$\lambda$5007 ratio ('ratio temperature'; $T_{\rm ratio}$)results in an underprediction of metallicity when temperature fluctuations arepresent. In contrast, using the unobservable 'line temperatures' ($T_{\rmline}$) that provide the mean electron and ion density-weighted emissivityyield an accurate metallicity estimate. To correct this bias, we demonstrate anexample calibration of a relation between $T_{\rm ratio}$ and $T_{\rm line}$based on a high-resolution (4.5 pc) RAMSES-RTZ simulation of a dwarf galaxythat self-consistently models the formation of multiple HII regions and iontemperature distribution in a galactic context. Applying this correction to thelow-mass end of the mass-metallicity relation shifts its normalization up by0.18 dex on average and flattens its slope from 0.87 to 0.58, highlighting theneed for future studies to account for, and correct, this bias.

We present re-processed flux calibrated spectra of 406 stars from theUVES-POP stellar library in the wavelength range 320-1025 nm, which can be usedfor stellar population synthesis. The spectra are provided in the two versionshaving spectral resolving power R=20,000 and R=80,000. Raw spectra from the ESOdata archive were re-reduced using the latest version of the UVES datareduction pipeline with some additional algorithms that we developed. The mostsignificant improvements in comparison with the original UVES-POP release are:(i) an updated Echelle order merging, which eliminates "ripples" present in thepublished spectra, (ii) a full telluric correction, (iii) merging ofnon-overlapping UVES spectral setups taking into account the global continuumshape, (iv) a spectrophotometric correction and absolute flux calibration, and(v) estimates of the interstellar extinction. For 364 stars from our sample, wecomputed atmospheric parameters $T_\mathrm{eff}$, surface gravity log $g$,metallicity [Fe/H], and $\alpha$-element enhancement [$\alpha$/Fe] by using afull spectrum fitting technique based on a grid of synthetic stellaratmospheres and a novel minimization algorithm. We also provide projectedrotational velocity $v\sin i$ and radial velocity $v_{rad}$ estimates. Theoverall absolute flux uncertainty in the re-processed dataset is better than 2%with sub-% accuracy for about half of the stars. A comparison of therecalibrated UVES-POP spectra with other spectral libraries shows a very goodagreement in flux; at the same time, $Gaia$ DR3 BP/RP spectra are oftendiscrepant with our data, which we attribute to spectrophotometric calibrationissues in $Gaia$ DR3.