Thanks for checking, Derrick! I suspected as much. After I saw your note, I glanced at the Illumina product spec sheet for the "Infinium OmniExpress-24 v1.2 BeadChip" and it lists 17,565 markers tested on the X-chromosome (my raw Ancestry data is consistent, but shows 17,605) with 683 tested at PAR (pseudoautosomal) loci. A little more digging indicates that, for the build 37 map, the OmniExpress-24 considers 62,321 to 2,697,868, and 154,939,018 to 155,236,747 to be the PAR regions. The same digging also indicates that, for genealogy testing, the companies routinely discard the PAR data even thought the chip supplies it; understandable: that potentially heterozygous data doesn't have any genealogical bearing and would just muddy the waters. But it's also unsurprising that a PAR marker or two will slip through into the published results with some frequency.
On the yDNA front, I also suspected my inclination to check the Y-SNPs shown with the autosomal results wasn't very original. And it wasn't. :-) Our own Dr. Turner even commented on it a couple of years ago.
Net message is that those tag-along Y-SNPs that are captured in atDNA tests really don't tell us anything. An FTDNA project manager compared the AncestryDNA SNPs to those in the ISOGG index and found that of the 885 SNPs he examined (the total shown in his AncestryDNA results), only 318 were in ISOGG's index. The remainder did list in the Ybrowse database (but that uses an NCBI full genomic dataset, so I'd fully expect all the reference clusters to appear there, just as they do in the "1000 Genomes Browser"). And of the 318, 256 can be listed as "relevant" to a haplogroup because they appear in the mix somewhere...but in general they're too far upstream to tell us anything without having the full complement of hierarchical SNP results. For example, he found 79 SNPs related to haplogroup E, 72 to R, 40 to I, and the rest had 19 or fewer SNPs.
Too, there seem to be issues with some heterozygous calls and false positive results in these atDNA Y-SNPs...I suppose understandable since the tests aren't for-purpose and don't go through the multiple sampling and quality control that a Y-SNP test would. So even if the SNPs tested were more directly applicable to yDNA haplogroup analysis, there would be questions about accuracy and consistency in doing so. On the bright side, though, that means I finished my weekend project by noon Saturday. ;-)
P.S. I'm done now, Pete. Honest.