Right, for matrix corrections to work, you at least need all the major elements. You also need something to relate background to the Th/LREE concentration and subtractions for the Pb counts for interference with Ylg and ThMz, and adjusting UMb counts for ThMg. We have not done age mapping in some time, but basically we used empirical Th/REE background (you can collect background maps too, although I think this just adds error to such minor concentrations for Pb and U) and interference corrections on the counts based on spot analyses, then applied Z A F factors also based on results from full major element analysis. All kind of crude, but where age domain differences are large enough, can make for a nice map. I would imagine with ProbeImage one could get the EDS spectrum to roughly quantify P and REEs to do this much more accurately. That way you could actually do real interference corrections and the MAN background might do well enough. Overall, you are better off using a single bkg value for a single compositional domain (where Th/LREE and Y do not vary much) rather than pixel by pixel values with large count uncertainty, so maybe a good way would be to assign some sort of count range of Th and Y to a single background value, and another range to another bkg value, and so on.