We are trying to establish the accuracy of an Eagle III µXRF with vacuum chamber and analysed 3 different glass standards embedded in epoxy resin namely BRU7, NIST 610 and NIST 620 at 35 kV and 150 mA with 31% datm. the results show much lower percentage of Silicon, Potassium and Calcium along with much higher percent of Sodium and magnesium while Aluminium values are quite acceptable. These standards were analysed with the same conditions in 2010 when they were bought and yielded much more accurate results. I wanted to ask if anybody has any idea about the reason of this discrepancy and how to correct it.