Rear-engined cars are quite different to the other formats due to the extreme rear-biased weight distribution. Most of the early Porsches were what we’d consider now a real handful to drive, particularly due to the rapid jump in engine performance and mass, and most of their advantages lay in the mastery of the lift-off -> power oversteer. They were particularly hairy under hard braking.
One of the issues here is that while the handling model for most cars within reasonably common limits is quite good, it’s hard to say how accurate they are at the extremes of power, weight distribution etc… If you compared this to another car at low speed without absolutely caning the throttle, does it feel similar or does it feel that the front is too light and there’s not enough friction in the front wheels? What’s the low-speed corner speeds really like? High tyre stagger will definitely make the car understeer more at low speed, but to really test what its limits are, since effective friction actually reduces when the tyre loses traction, you’ll have to find the balance where there’s no understeer to really test how fast it can corner.
As for high speeds in a rear-engined car, I’m also at the point where I can’t tell whether it’s the model or it’s my tuning or a combination of both that limits me from being able to consistently tune an MR car that can turn both slow and fast. As a general rule the stiffer your suspension the twitchier and more responsive it is but depending on the geometry and the tyre and the loading, too stiff and the wheel will either be overwhelmed or lose contact with the ground and that will limit grip (front) or make the car very nervous (rear). Not going overboard with rear swaybars and rear spring rate helps keep the car a bit slower and more predictable (within reasaon).
I haven’t worked out the relationship between the downforce and cornering performance or the high speed graph in Automation. IIRC downforce in Beam has been iffy up to this point and my impression is that it’s still not fully working yet, or the model is too simple and reacts funny in certain situations. I’m no expert on the actual code, though, so don’t take my word for it, but what I do know is that the values given in Automation and the behaviour at high speed in Beam still don’t match sometimes.