Hi Theo!
A couple of comments... First, just looking over the data table you attached. First, seems like the TiO2 contents of points 10 and 11 are high even for end member ulvospinel, so must be some exsolution with ilmenite or rutile (?) if this is supposed to be titanomagnetite. Boundary fluorescence of Ti in fine Fe-Ti oxide mixtures can be problematic too.
Second point, maybe a little out there, but magnetite-titanomagnetite can be magnetized, having picked up permanent magnetization especially if there is maghemite or other impurities that might increase the magnetic retentivity. If they have been polished on a system using magnetic disks, that could do it too. We don't do magnetite very often, but we have noted beam deflection from time to time compromising x-ray focus, and we wonder if the current flow from the beam could actually induce some of this effect under the right circumstances. We have analyzed tetrataenite which is even more problematic, with some very dramatic magnetic field effects, but we could compensate by storing a beam deflection (in the setup) to get the beam back into correspondence with the position during calibration. Very tedious because the fields would vary grain to grain. So, anytime we think about probing ferromagnetic minerals, we consider beam shift as a possible variable. As you note, your ilmenites seem fine, so appealing to MAC tables is unlikely. Anyway, maybe beam deflection is a possibility?