Data submitted over the past 25 years to GeoPT, the highly successful proficiency-testing programme for the geochemical analysis of geological materials, organised by the International Association of Geoanalysts, provide a valuable resource that permits detailed investigation of contrasting results associated with different sample preparation and measurement principles. Highlighted issues include the following: recurring problems with the dissolution of the refractory minerals zircon and chromite, which produce a large dispersion in data obtained when acid digestion is involved; issues related to different XRF sample preparation methods, whereby a significant divergence of pressed powder pellet results compared with those from fused glass discs is observed; high relative dispersion of data both at low mass fractions, and those higher than normally found in silicate rocks, due to incomplete method validation, in particular due to an overconfidence in estimating reporting limits and to the employment of limited working ranges. In addition, an example for Sr in an ancient pegmatite with extremely high Rb abundance is presented, where ICP-MS results amount to only to a third of the XRF results, severely underestimated due to the disregard of the radiogenic ingrowth from 87Rb decay. Recommendations are made both for improving data quality and the selection of test materials for future GeoPT rounds.