When making the single analysis (not multiple) with the sample (without even taking it from the box and putting back), predictions are different, deviation is within 5% - why is it so?
Regarding the deviation that has been recognized performing single analysis on samples, it is quite normal in case of material not homogeneous (as it is for example hay). As a matter of fact, in case of single analysis, the surface of material scanned is limited. Repeating the analysis, different part of the material will be probably scanned and not-homogeneous ingredient surface will provide significantly different spectra. This is the reason why the suggestion is to perform Multiple Analysis instead of a Single one.