Metal–silicate partitioning studies performed in high-pressure, laser-heated diamond anvil cells (DAC) are commonly used to explore element distribution during planetary-scale core–mantle differentiation. The small run-products contain suitable areas for analysis commonly less than tens of microns in diameter and a few microns thick. Because high spatial resolution is required, quantitative chemical analyses of the quenched phases is usually performed by electron probe microanalysis (EPMA). Here, EPMA is being used at its spatial limits, and sample thickness and secondary fluorescence effects must be accounted for. By using simulations and synthetic samples, we assess the validity of these measurements, and find that in most studies DAC sample wafers are sufficiently thick to be characterized at 15 kVacc. Fluorescence from metal-hosted elements will, however, contaminate silicate measurements, and this becomes problematic if the concentration contrast between the two phases is in excess of 100. Element partitioning experiments are potentially compromised; we recommend simulating fluorescence and applying a data correction, if required, to such DAC studies. Other spurious analyses may originate from sources external to the sample, as exemplified by 0.5 to >1 wt% of Cu arising from continuum fluorescence of the Cu TEM grid the sample is typically mounted on.