Is actually quite a valid point, and an interesting discussion idea. Because there are 2 sides to that. Just to take a quick second - which is more immersive, something looks stunningly like an F16, but has no connection to how the real thing behaves and none of the switchology, or something that looks "dated" but 'flies' just like the real thing and has all the systems implemented? (not arguing your point or ideas, just taking a quick side step down that discussion path

)
Well i tend to think i am quite balanced but then i cannot be perfectly unbiased about myself.
I do think substance is needed. For example i never liked Doom3, which sported the best VG graphics for some time, because it was only graphics. While i praised and loved HalfLife2 which what lacked in graphics added in feeling, story-telling, immersion.
That said what can i say about Orbiter?
It looks like what it is: a computer program. And all the difficulties inherent to it (from "simple" scenario creation, to vessel implementation) while adding to the elite-feeling, do substract from easiness and transparency (imo). A kind of "virtual traffic" for example would really add to the "alive and real" feeling.
But here what is the most basic point?
Simply put: technology evolves and gives us new possibilities, why not exploit them?
I can see you saying that graphics are not so much important when flying to neptune. Bus is flying to neptune all there is to it?
Or maybe takeing off at night and arriving at the iss at dawn may be part of it too?
By the way even orbiting around neptune, with nice materials, lighting effects and high dynamic range, could be spectacular...
And why not add a beautifull view to the satisfaction of having achieved orbit after days of carfull manouvering?
Edit to add - I would certainly love to see some of those "add-on" enhancements become part of the core, and perhaps a little better implemented by being (part of the core). Most especially 3D terrain. But the biggest thing I'd love to see in any Orbiter development, is collision detection.
You know?
Mesh intersection detection could be one of the most processor intensive tasks ever. More so then graphics.
No surprise it's usually done with "bounding geometry" with shape approximation being fairly gross.