Fundamental parameters methods of calculating elemental concentrations from x-ray fluorescence (XRF) data put stringent demands on the accuracy of XRF intensity measurement. Two significant but rather unobvious sources of intensity error are inaccurate deadtime correction in energy-dispersive x-ray fluorescence (EDX) systems and errors caused by certain combinations of sample thickness and geometry. A method of evaluating the accuracy of an EDX system's deadtime compensation is given. In our system, a nearly constant error was found over a wide range of average x-ray photon energy. Accurate deadtime correction was obtained by adding a mathematical correction to the partial compensation by the built-in system electronics. Problems related to sample thickness and geometry can result when the effective x-ray path length is longer than the geometrically possible path length (high energy analyte radiation and light element matrix). In such samples, a known sample thickness less than the geometry-limited thickness must be used for accurate intensities. A method for determining when this is necessary has been devised. A useful empirical correction for absorption by Mylar windows is also given.