It looks like system "memorized" target speed and path, and adjust missiles to the target, even when not tracked on distance...

Subsonic target UPR point is closer to the target. It can be less than 4 degree, depending on the P.
Supersonic target UPR point is further to the target. It can be more than 4 degree, but the system allows the missile max 4 degree.
When the missile is flying at less than 4 degree, than it is directly goes towards the UPR point.
When the missile is flying at 4 degree, than the UPR point is at probably more than 4 degree, just the system limits its lead angle.

This guidance method is simply UPR, just adding a Constant "K" into elevation, depending the distance of the missile to the range boresight.
If the missile - boresight distance is larger than 10km, it adds 1.2 degree to the elevation boresight.
From 10km, it gradually reduces this Constant to 0 degree. (at 5km it is 0.6 degree)

Seems that in ver. 923.0, in jamming enviroment, K became like T/T with 1.2 degrees, and UPR is just like T/T. Still they react on distance boresight. If UPR is based on real target speed, (sub- or supersonic) of the target and Parameter (this should be important only for the azymuth lead) against SNR,I believe - based ot what I've learned from this topic that UPR/K behavior in 923.3 is WRONG due to denial of speed and parameter information because of active jaming and must be corrected.
I would be glad if anybody shares his own observations.