It seems to me that because the three plate solved positions are reported as unrefracted JNow positions the pole position derived from them will also be unrefracted JNow.

Assuming that is the case then the best way to get to a different pole would be to correct for refraction in the JNow coordinate system. I would be less confident correcting for refraction in the observations because it will vary according to the altitude of each observation.

It's pretty good when you are getting close enough that you are worying about what pole you want to align to.