This work presents a structured investigation whose objective is more reliable efficiency measurements at a power-circulating rig. The German Research Association for Drive Technology (FVA) has established a method to measure gear efficiency for this rig concept, assuming that each of the two boxes dissipates half of the amount of power delivered by the motor. An experimental investigation was designed and conducted in order to verify the impacts of transmission boxes with different power losses over the dynamic behavior of the power losses inside the loop. A power-circulating rig was conceived with torque measurements inside the power loop and with a system able to induce controlled losses independently on each transmission box. A full factorial design was used to identify significant factors and interactions and this experiment plan was applied under two conditions: with and without locked-in torque. From the experiment results, it was possible to establish the power flow direction and draw conclusions about the relation among the transmission boxes' losses. In the test rig used for this study, it was shown that losses on boxes depend on loss variations on other boxes belonging to the power loop, mainly on those located upstream with relation to the power flow direction. Remarkably, one of the boxes was less susceptible to other boxes' loss variations. A hypothesis for the determination of the less susceptible point to variations in other points of the loop was proposed based on the literature review and the experimental observations.