There is no simple answer to this question. So you are dealing with a situation where the results of the analysis of the elemental composition of a specific cement sample obtained by different methods differ significantly. Why do you think the differences are significant? Do these results have assigned uncertainty values? And would you like to know which method is generally more accurate? Theoretically, classical methods are more accurate, because XRF techniques are calibrated and checked with reference materials, the certified values of which were established mainly by classical methods. In a case like yours, this can only be checked with additional analyzes performed by another analyst in a specialized laboratory.
X-ray fluorescence (XRF) is commonly used in cement plants. It is also a primary means of controlling the composition of raw materials, the raw feed, as well as clinker and cement. XRF is very effective at ensuring the provision of compositional data for controlling almost all stages of production and also assessing the final product. However, the choice of method should depend more on the ease with which the results can be independently verified.
I agree with Zbigniew Jońca and want to add further
Although, theoretically these are accurate, extra care is required using classic methods. Too many steps involved in classic methods and some of the techniques such as filtration (especially for the aluminosilicates, or Calcia-aluminosicates or siliceous) may induce errors. Even reproducibility of results is less.