TSVs produced of windowed pi for Dmel and Dsim do not use "X", "2L", "2R", "3L", and "3R" for their scaffold name columns, so they are filtered out in R. The issue is different for Dsim than Dmel, since the FlyBase Dsim assembly uses "Scaf_*" for the major chromosome arms, whereas the Dmel issue is inclusion of text after an initial space that is not omitted by my scripts.
Dmel would be fixed by ignoring any text in the scaffold header after the first whitespace.
Dsim would only be fixed by renaming the reference and pseudorefs, but that's a rather silly solution. A perhaps more long-term solution would be to use our Dsim w501 v3 (PacBio) assembly, but that assembly and annotation are for a different manuscript, plus would require a complete redo of all constraint analyses.
Leaving this open as a known bug, but not planning to resolve it anytime soon.