Even a fresh 9V battery has a high internal impedance. And they droop quickly with a significant load.
Using a good power supply, well filtered and regulated would be requisite. Also measure the key-down current and the power input to the final amp: determine efficiency. Then there is the also mentioned power meter. The QRPometer is a 2% accuracy meter, as I recall.
Otherwise, you might get good results with everyone using the same measuring method; a peak detector or an RF probe for instance and account for their associated tolerances. A 10 meg, 1% DVM would help too.
As Terry sez, from his QC experience and mine too, measurement can be a tricky business. Especially if you don't have traceability to a "gold standard"...
Chuck Carpenter, W5USJ
EM22cv Rains County, TX