I believe the NDL is based on GFHi. If GF99 is the real-time indication of the CTC's pressure status then at an NDL = 0 the GF99 should be, very close to if not, GFHi. The SurfGF will be much higher as it represents an instantaneous surfacing. So, SurfGF does not take into account offgassing at predicted ascent rates.
the video on shearwater covers it really well. If you comply with the rates etc in the software then gf99 equals gf hi at time of surface. GF high controls the assent protocol through NDL and established ascent rates. gf hi sets a time where those 30 ft per minute will get you to the surface at the designated gf hi. GF99 is the current GF value and it decreases as SI progresses. I think I have that understood correctly. now if you extend your safety stop then the gf99 upon surfacing will be less than the setting gf hi. as it should be.