First thing first it is pretty stuiped the smart thing to do would be to add a 1/60 second lag and then do linear interpolation through the model animation and position all though this might not be correct in the physics it would be damn close enough.( it would be more accurate then drawing it at the same spot again in 99.999% of the cases anyway ).
Benchmarking you got a few options
You got a system benchmark which disables any waiting between game ticks that lets you run the game faster then possible. ( best but prolly gonna be cpu limited )
2nd option is to let it rerender identical frames it wouldn't surprise me at all if carmack did this because it would be impossible for people to tell how many frames were REALLY rendered nvidia could say it rendered each frame 100 times in the drivers and yeah we couldn't prove they lied. ( or they could grow morals but for they amount of money they gonna make I'd prolly do the same thing
3rd option is what they should have done for game play above.