Your assumption regarding the Audi/Pugeot diesels is correct.
The blue data largely represents derivatives of production engines. Since, the data scatter is rather large, because it covers a rather general range of engines, it can be assumed that P2 engines receiving some level of race modification, would preferentially be represented by the lower half of the blue data range as you have showed with the "green' line. For your project, the slope of the green line is the important consideration i.e., how is inertia expected to change as a function of increasing torque for the P2 engines.
Perhaps with more data representative of the "blue" definition, and a narrower filtering of the engines types/makes, the blue trend line could be replaced by other lines such as your green line.
The pink line represents the limited data that supposedly depicts the inertia/torque characteristics of ultra-high output, purpose designed and built race engines.
Anyway, the subject is interesting.....hopefully, we'll find addition data that can make the chart more useful.