I'm no expert in this area, but I take it you trying to ascertain the geometric height of an aircraft with the information available (which unfortunately does not include GPS altitude).
So you are starting with a barometric ADS-B readout, and applying local QNH to give a geopotential height (ignoring transition level, as you have stated - it's irrelevant), which you then convert to geometric.
That sounds fine, but you could probably get more accurate than this.
Firstly, QNH will be accurate when an aircraft is near the ground, but sea-level pressure variations may be of little use for an aircraft at a higher level. You may be better off looking for upper air charts. High level highs and lows do not necessarily correspond to low level highs and lows.
The next step then needs to be temperature correction. In cold weather the aircraft will be lower than it is broadcasting, and vice versa. Once again, sea level temperatures will be of no use (it may be ISA+20 on the ground but ISA-10 at high level), and in fact the QNH includes a temperature correction for ambient, but only valid for the height of the source of the QNH.
This page may help. Again, air charts are available for temperature. At heights above 36000 feet, ISA assumes a zero temperature lapse rate (stratosphere). I actually don't know if aircraft altimeters are calibrated to correct for this, but I doubt it. Assuming that they don't, aircraft above 36000 feet is ISA conditions (-57C) are probably slightly higher than they think they are, so your temperature corrections probably need to disregard the tropopause also. Perhaps someone can help with this.
Anyway, the point is that actual temp/pressure corrections for the altitude of the aircraft will give you the most accurate geopotential height that I can think of.