My two cents...
First of all, I'd like to chime in and tell Rob and Nova that their stuff looks great! It's obvious that you guys have a lot of computing horsepower behind you to be able to do what you do...not to mention an unlimited love for the Constitution-class design (as I do).
I can relate to the production dilemma of time versus money. I deal with it on a daily basis in my job, so I understand the problem of wanting to do more (or better) but being unable thanks to schedule or budget constraints.
As some other posters in this thread have mentioned, I also dabble a little in both physical and CGI modeling, and yes, many of the techniques are the same. When it comes to what people call "photorealistic" CGI images, though, I have to ask one question: To what are you referring?
There are many kinds of "photos," from the sharpest, high-quality portraiture to happy snaps somebody takes with a disposable camera. And there are almost all kinds of lighting involved as well. To say that something is "photorealistic" implies more than just saying the scene was shot with a camera. When you involve filmmaking techniques like bluescreen compositing, "photorealistic" takes on a whole new meaning.
Like, for instance, what kind of lens are you using? What type of film stock? Or, if it's video, what's your focal length? If you're rendering for TV or film, you also have to take motion blur into account, and it's different between television (30 fps) and film (24 fps). Also, film has more color saturation than videotape. There's lots to take into account when trying to create something "photorealistic," and it's no wonder that folks like Rob run up against that brick wall each week. To do it that way takes more money and time than a weekly TV show will allow. That he's able to do it at all simply speaks to his skill (and the skills of his co-workers) as CGI artists.
I'm working on a Big E of my own using Cinema 4D (on a laptop, no less), and I've discovered the pitfalls of trying to do "photorealistic" renders. For instance, all of my texture at this point (with just a few exceptions) is bitmapped, which makes for a pretty picture...when you're looking straight on at the subject. If it's turned to an angle, it blurs out, and destroys any illusion of realism. But modeled-in textures (such as the registry pennants and things of that nature) add to the size of the mesh, and when you're dealing with a Celeron processor and 256mb of RAM, you don't have the luxury of doing more than making sure your shapes are the correct size, contour, and relationship to each other (which I'd like to say I've done pretty well, IMNSHO

). It's given me a new appreciation of what folks like Rob and Nova go through to create what they do, and it's also taught me a lot about physical modeling.
I do have some ideas about a modeled-in shuttlebay, though...
