@Austin

Regarding point 1: Your friend can run it himself if he wants to verify. I got the result by BLASTing the sequence provided in the article (VIVMLTPLVEDGVKQC) against SwissProt's db of Bacteroidetes, Firmicutes, and Gammaproteobacteria (TaxIDs 976, 1236, 1239) because I was looking for human gut bacteria. Standard parameter short-seq BLASTP on blast.ncbi.nlm.nih.gov. Legionella's tyrosyl tRNA synthetase is the first hit.

Two points regarding point 2:
1) The E value—i.e. the number of hits of this quality you'd expect by chance, in a database of this size, is 0.005. This is more relevant than what "seems great" to your friend.
2) The "3D structure" is not relevant, because of how antigen presentation to T cells works; the MHC presents a linearized fragment of the protein. Please see: https://en.wikipedia.org/wiki/Major_histocompatibility_complex#/media/File:MHC_Binding_Diagram.png

On point 3: There is not sufficient data on non-Legionnaire's Legionella exposure to compile this kind of map at the resolution where it would be useful.

On point 4: These would not be expected to align perfectly, as:
1) the top graph measures incidence of Legionnaire's disease, which—as discussed—is not 1:1 associated with exposure to Legionella.
2) Legionella exposure peaks at different times of year in different locales, as evidenced by the differences between cohorts in that study (green vs blue vs red lines), which would "fuzz" any associations when comparing across populations (e.g. German cohorts, top, vs. US cohorts, bottom).
3) there might be some temporal offset between exposure and onset of disease.

Would appreciate a review from a friend with immunology expertise