Occasionally suggested yet rarely performed X-ray fluorescence (XRF) spectrometry of fluorine seems to fail systematically in yielding reliable quantitative results for rocks and soils. Repeated analyses reveal continuously drifting fluorescence intensities for fluorine, boron and chlorine. Typically, an increase, but in few cases also a decrease, over X-ray exposure time is observed. For instance, fluorine concentrations in a soil standard appear to increase steadily from below the detection limit in the first run to nearly 850 mg/kg F more than 10 h later in the last. In contrast, cryolite is characterised by drastically decreasing intensities for fluorine. Although fluorescence intensities may be affected by preparation methods, specimen surface conditions and dynamic contamination, it is shown that none of these influencing factors is responsible for the observed trends. In fact, there is evidence that X-radiation impact mobilises fluorine, boron and chlorine. Diffusion of radiolysis products towards the specimen's surface as well as the kinetics of adsorption and desorption or chemical reactions are believed to control the analyte concentration in the analysed layer decisively. Furthermore, during analysis, the latter is altered by considerable losses of binder or fluxif applicable thus enhancing XRF intensities of boron and fluorine because of reduced absorption. In any case, signal stability appears to be limited by insufficient sample and specimen stability. It is concluded that for many soil and rock samples, XRF spectrometry is inappropriate to quantify fluorine, although the crucial obstacle is neither the analytical method nor the spectrometer sensu strictu.