I believe he just meant variability between runs as in path/movement/shooting,/ai, etc. Its not the same frames going through each run...
...but such as it is when there isnt a built in benchmark. You can only do the best you can and try to repeat as best you can. That said, I know the Dirt series benchmark isnt the same each time (same track and weather, but finishes in. Different.place, like 4th then 2nd, etc). and it is integrated/canned (but still within 1%). I want to say F1 2018 is too? There's a couple.