The reason we use in-game time in the first place and not real-time is because different computers run the game at different speeds. If you compare a lot of different people's runs with similar in-game times, you will notice that they are often wildly different in terms of real-time. For example, Goomba runs this game very smoothly and loses almost no real-time to lag, whereas many others such as myself and f1 have computers that run the game more slowly. To prove this to yourself, try comparing real-time speeds on straightaways (where neither person is getting boosted from behind), since they can often show you how one person starts to slowly creep ahead of the other while the driving is basically the same. The effect is subtle but should be noticeable, and over a whole run that subtle gap in real-time speed adds up.
The method you tried to compare between computers doesn't work because of how the game progresses the cutscenes. In particular, the zoom-out at the end of the game will run for the same time on different computers regardless of framerate because those sequences are governed by the system time of the computer (after your computer's internal clock advances 5 seconds, the game moves on, regardless of framerate). This is much different from standard gameplay, where the physics is capped at 30fps but can dip below that if you are experiencing lag, resulting in RTA timelosses.
The in-game timer accounts for differences in lag because it measures the number of physics steps taken, not the real-time speed of the run. If a computer were to run the game at an average of 28 fps, it might take them 180s RTA to go a certain distance, while a computer running at 30fps would get there in 168s. However, in both cases, the game takes the same number of physics steps to move between those points, so the times are able to be compared.
You can also hit me up on discord if you want to talk about this more. Our semi-official server that we used to address timing concerns in the past is here: https://discord.gg/u2YK4yr
