A new empirical calibration of the (B-V) index versus the three fundamental stellar parameters T(eff), (Fe/H), and log g is presented whose range of validity is adequate to analyze the main sequence and turnoff regions of most of the stellar cluster systems. Significant systematic differences are found between this new calibration and previous ones based on model atmosphere predictions. On the one hand, solar abundance models are systematically too blue for T(eff) between 8000 K and 6500 K and too red for T(eff) between 6500 K and 5000 K. On the other hand, it appears that current atmospheric models are underestimating the variation of the (B-V) color due to metallicity. When the new calibration is used to compare recent high-quality color-magnitude diagrams with the revised VandenBerg isochrones, very good agreement is found for the metal-rich or moderately poor clusters, but discrepancies are found for the most metal-poor clusters.