Lunar glass beads, including volcanic and impact origins, have become key targets for in-situ U-Pb dating using secondary ion mass spectroscopy (SIMS) to trace the thermal evolution and impact history of the Moon. However, the accuracy of SIMS U–Pb dating results is questionable without careful evaluation of matrix effect because of the largely variable major element contents of lunar glass beads which are also significantly different from the commonly used glass standards. In this study, we used seven glass standards with large variation of compositions, including three MPI-DING glasses, two USGS glasses, ARM-3 glass, and LMG-4 glass with composition simulating the Chang’e-5 lunar soils, to investigate the U-Pb fractionation behavior in SIMS analyses. Our results revealed more than a threefold variation in Pb+ ion yields, which positively correlated with the total FeO content (FeO(t)), and an eightfold variation in U+ ion yields, which negatively correlated with FeO(t). The relative sensitive factor (RSF) of (206Pb+/238U+)/(206Pb/238U) shows a variation of over 28-fold. Therefore, the traditional U/Pb calibration method based on only a single standard is insufficient to correct such a significant matrix effect. We propose a calibration protocol based on a linear relationship between ln((206Pb+/238U+)/(206Pb/238U)) and ln(UO2+/U+), or alternatively, ln(206Pb+/238U+/206Pb/238U) vs. ln(UO+/U+), and ln(206Pb+/UO+/206Pb/238U) vs. ln(UO2+/UO+). This method requires at least two glass standards with a large range of FeO(t) to effectively correct for matrix effects. Additionally, we recommend revised 206Pb/238U values for those glass standards that lack values determined by isotope dilution thermal ionization mass spectrometry (ID-TIMS), based on the observed fractionation rule in this study.