What I don't understand is... why your test setups don't try to mimic the thrust in real life ?
i.e. an 30cm horizontal arm, articulated at one end, and fixed to the ground with a digital scale like this
The arm should be placed at bench margin, enough height to minimize ground effect.
Having known the motor/arm/esc/wiring weight, give it a spin, and, when it hovers, you have a first reading, power to hover xx grams, own weight.
Then start to increase throttle, and have readings at various levels.
An important reading for me would be at half throttle, this way I have the optimal AUW for that motor/prop combo, of a full equipped frame.