We have studied the performance of global chi^2 fitting of low-resolution X-ray spectra in retrieving intrinsic source parameters, with emphasis on the coronal metallicity. The study has been conducted by fitting large numbers of simulated spectra with known characteristics, and studying the distribution of best-fit parameters. We have studied the behavior of the LECS detector on board the SAX satellite and the SIS detector on board the ASCA satellite. The fitted source spectra have either two discrete temperature components or a power-law temperature distribution, with metallicity variations modeled by a single global abundance parameter. The model used for the fitting has always been a two-temperature one, with global varying abundance, to explore the influence of the a priori ignorance of the actual temperature stratification in the source being observed. The simulations performed explore the influence of varying statistics in the observed spectrum (spanning a realistic range of values) as well as the effect of varying the intrinsic source metallicity, with values in the range 0.15--1.0 times the solar value. We find that the source metallicity can be retrieved within few tens of percent from ASCA/SIS spectra of typical signal to noise ratio, and within few percent from SAX/LECS spectra at the same signal to noise ratio. However relatively small uncertainties in the detector calibrations and in the plasma emission codes are likely to potentially cause large systematic off-sets in the value