Grinchy: that's why Vapor went a bit furthier to reduce margin of error with his multiple mounts/dropping best&worst results/and using average results, and imho his results can represent what user may expect in most cases and makes relative TIM performance be comparable even with human aplying.
(of course, if other points get taken care of aswell, like constant ambient temps and sufficient cure time for all pastes).
And imho no need to use artificial test-bed and artificial aplying methods .. while they may actully better represent real capabilities of TIM interface, it won't represent what average joe might get aplying himself on real cpu/cooler where big role alongside TIM's °C/m W plays viskosity and alike properties meaning simple thing "can this paste be aplied by mortal effectively, or it's superrior thermal transfer properties will be nullified by it being PITA to apply properly"