I'm just wondering what everyone's major distaste to using props is? I guess it's the fact that it's too easy to be seduced by the ability to use the same model again, especially if someone else has already made one.
The thing is, the props give you a lot better performance. I'm not sure if Hammer has this ability, but if a structure is purely detail, convert it into a prop for several reasons. Props run faster because all the geometry for them can be sent at one time since it is uses fewer shaders/textures, making it so the renderer doesn't have to switch as often. It's not how much you send per frame, it's how many times you send per frame.
Also they are much cheaper to store. You only have to store the model/prop once in memory, and then just a transform matrix(4x4 matrix, 32bytes per element), and some other unique identifiers for it that are low memory for each time you use one. Also with them you usually have control of the collision bounds of them as well. That transform matrix gives you a lot of control over the model as well, you can move, scale and rotate as much as you want on any of the three axi'(sp?), allowing you to drastically alter the look of the object. Infact we had a really cool effect where we turned all our models large and flat along the Z axis, making things look like Paper Mario, or putting animated models onto a movie projector.
Yes I know it takes a bit of time to learn out to use a modeling program, but it's really worth it, especially with the tools available for making complex geometry. I think to help you need to promote modellers to come here, or make friends with those over at Polycount and create the ThePolySnarkCountPit!!!!