The above circuit and thesis is reasonably OK. A couple of things not included in the above engineering are:
- loss of course in the line/circuitry leading to the above measuring circuitry, including to the two resistors.
- The thermo properties of the two resistors being heated by the signal-under-test and the DC signal. You have to, of course, also include the TRANSIENT thermo properties of both (I never did trust DC, including the wiring to/from the power supply.)
By the way, I believe your above circuit, including it basic premise, is covered by a patent taken out by Norm Dillman while at HP.
Myself, I used similar circuit approaches as this to provide thermal biasing stability on the many class B power amplifiers I designed for GE Audio Products back in the 60's.
But you have a nice circuit as well as a nice paper describing it and your engineering.