2002.07.05 01:23 "Re: Photometric interpretation for grey-scale JPEG", by Tom Lane
I suspect that they were captured as colour images in YCbCr and then downsampled by simply discarding the chrominance components.
The TIFF writer probably didn't bother changing the PI, regarding YCBCR as an accurate description of the colour model given that the samples per pixel value was 1, implying just the Y channel.
I think it's nothing but an error. We've seen plenty of past examples wherein JPEG-encoded TIFFs had a silly choice of PI; that's exactly what TTN#2 is trying to warn against in the passage under discussion. I'd bet lunch that the software author had no such coherent rationalization in mind, but merely forgot to consider the issue.
I don't think the spec should be relaxed to retroactively legalize images that are clearly broken. TIFF is already a nightmare for readers because it bends over backwards to let the writer do anything it darn pleases. There is no functional justification for being permissive on this particular point.
regards, tom lane