Oystein: (Provided that pigments in swollen MEK/Tnemec chip are still distributed more or less evenly and almost all XEDS maps are more or less unrealistic.) What is the main difference between Al and other elements measured in Fig. 15 (Fe, O, C and Si)? According to XEDS of MEK chip before soaking in MEK (Fig. 14), Al is by far the
rarest element among them. This is of course just "optimal" for mighty thermitic weapon

, but anyway... Could it be that a strange XEDS map of Al
(illusory concentration of Al at chip's "slopes") is somehow connected with the very low concentration of aluminium in the chip?
Difficult to say...
The Almond has several times explained that EDS spectra can be bitches, as some frequencies may be more muted by absorption by matrix material than others, and that this depends on the matrix material. In particular, he said, Nitrogen is notoriously difficult to detect in such materials. So who is to say how different peaks for different elements are affected by the unidentified matrix material, and the geometry of the probe?
Imagine the Al was found only on the surface of the red material, in small concentration. Then looking perperndicularly at the surface, you have only few Al-atoms per pixel, but looking on edge, down a steep slope, you have more per pixel (which is a square of some spatial angle). This would be a theoretical explanation for why we get lots of signal on the slopes, and little in the "plains". If the Al were evenly distributed throught the 3D depth, inclination shouldn't make such a difference.
Unless it's not the Al that has a gradient perpendicular to the surface, but the matrix material: Perhaps the matrix material strongly blocks the Al signal, but the MEK soak has changed the surface of the chip such that it doesn't block Al-rays as thoroughly? Then we'd see the same effect: More Al-signal get through per pixel at the slopes.
So here I have two different explanations; none particularly probable, if you ask me, but also none outrageous, and none ruled out. It boils down to speculation.
Still, as you noted several times, Al signals in Tnemec red primer should be accompanied by some calcium signals
(since only source of Al should be calcium aluminate). I just read in
Wiki that there is a whole "spectrum" of such aluminates available and some of them contain only one Ca atom per 6 Al atoms, which might perhaps explain missing Ca in Fig. 17 at least partially (if this chip is Tnemec). But I know, this is more "wishfull thinking" than acceptable explanation

A completely "flat line" over ca 2 KeV in Fig. 17 is anyway rather "suspicious", as you noted.
Fig. 14 shows a very high concentration of Ca - which other ingredients of Tnemec contain Ca? Are there Ca-Silicates? Carbonates?
Jones and Harrit aren't totally wrong when they are concerned about the effects of possible surface contamination before the MEK-wash. They haven't demonstrated it, so it is not valid of them to simply dismiss all the Ca and S as contamination, but it is also not implausible that some of the signal in Fig 14 is due to contamination, and gypsum surely is one of the more plausible candidates.
But
as I have shown, only a part, at most half of the Ca, can be explained with gypsum contamination. Then Ca is still much more plentiful (by a factor of 3 and more by atom count) than Al, so I'd put more money on Tricalcium aluminate (Ca:Al ratio 3:2), which WP describes as "
...the most basic of the calcium aluminates. It does not occur in nature, but is an important mineral phase in Portland cement." (Although Monocalcium aluminate would be the best candidate, with Ca:Al = 1:2)
Of course, the spectrum of Tnemec,
as shown by Jones in Sydney, has a very dominant Ca-peak and a small Al-peak, and an even smaller S-peak, so there has to be another main source of Ca in Tnemec, other than aluminate. And I suspect Jones picked his graphs such that Ca remains hidden.
This whole section on the MEK-chip in the Jones paper suffers from a combination of lack of focus and plan, inexpertly done work on the machinery, cherry-picked results, speculative dismissals and, possibly, altered data. Tnemec is still the most plausible explanation, but I am not as sure about that as I am about chips a-d being LaClede.