The issue here is that time loss can vary, even for the same car / track / game combo. If the benchmark is recorded during a refuelling stop but a few days later you race the same combo but only need tyres this time, it'll use the old and inaccurate benchmark.

I can't decide whether this is an edge case or not