Dale and I have published a paper in J Mat Sci (https://rdcu.be/dWZxp) evaluating the accuracy of low beam energy analyses using standards and EDS. The results are pretty encouraging.
"Testing the accuracy of low-beam-energy electron-excited X-ray microanalysis with energy-dispersive spectrometry"
Very nice article I enjoyed reading it. I am very interested in your analysis of the SiC and Si3N4 which as you pointed out had raw analytical totals further from unity than you might typically expect. We had seen the same thing in our own 5keV analysis of SiC (although we saw totals>1 (more like 1.1-1.2 range)) and were somewhat thrown off by it (re-ran analysis dozens of times under as many different scenarios as we could think of to mitigate whatever problem was going on). We ended up using the k-ratios obtained from DTSA-II and using badgerfilm to do the PRZ corrections. Interestingly the xphi model in badgerfilm gave analytical totals of 0.997 and the weight percents came into what we expected for the material