The determination of low boron concentrations in silicate glasses by electron probe microanalysis (EPMA) remains a significant challenge. The internal interferences from the diffraction crystal, i.e. the Mo-B4C large d-spacing layered synthetic microstructure crystal, can be thoroughly diminished by using an optimized differential mode of pulse height analysis (PHA). Although potential high-order spectral interferences from Ca, Fe, and Mn on the BKα peak can be significantly reduced by using an optimized differential mode of PHA, a quantitative calibration of the interferences is required to obtain accurate boron concentrations in silicate glasses that contain these elements. Furthermore, the first-order spectral interference from ClL-lines is so strong that they hinder reliable EPMA of boron concentrations in Cl-bearing silicate glasses. Our tests also indicate that, due to the strongly curved background shape on the high-energy side of BKα, an exponential regression is better than linear regression for estimating the on-peak background intensity based on measured off-peak background intensities. We propose that an optimal analytical setting for low boron concentrations in silicate glasses (≥0.2 wt% B2O3) would best involve a proper boron-rich glass standard, a low accelerating voltage, a high beam current, a large beam size, and a differential mode of PHA.