If the set of filters you are using, regardless of where they came from or what material they are made of, are successfully removing all of the extraneous higher wavelengths above the sodium D-lines doublet at ~589nm, then the only other immediate explanation is that the paper emulsion sensitivity is bleeding over to some degree into that wavelength. Either because it was designed to do that, or it has aged to do that.
Or something else silly is going on. Like, say, a leaky paper safe. When I did my testing I did take new sheets of fresh paper directly from the box it came in while in total darkness, then held them in a clamshell safe that I also occasionally use to store loaded film hangers, while the Duplex warmed up. I don't really think this is it, but it is a very slight possibility.
I am now looking for some time here to go down to my darkroom and also repeat the CD wavelength check. Maybe tonight.
Perhaps my color vision is slightly different than his and I missed the very faint residual green line that 'Sundowner' observed. Since I tested on Kentmere Bromide, which should be blue-sensitive only, I would still have seen no fogging if, as he observed, all of the blue was removed.
Tests on VC paper (which I haven't done under the Duplex) would be a different story however, if there was some minor residual green remaining. And I do note that if one looks at the Rosco transmission chart for the #19 filter, there IS a very slight transmission of greens. It looks to maybe only be 2-3%. But it's definitely there, if the chart is accurate.
For the record, of late I've been using only the red LEDs because I got into working with some Slavich graded bromide, and that paper is marked as red-only safe. Then when using MGIV I've just continued with those LEDs.