The north and south daylight cables are sited as piercing the roof at a point 100 km from where
they meet the ground. If we use this intersection to base the lower extents the windows on, again
we get a nearly perfect match to the given art - hot damn, consistent math and art!
Now the mirrors are trickier, for me at least. Took me a great while to figure them out. Still
don't know if I have, but now I have the model matching the book art better than I've ever been
able to.
My initial efforts looked nothing like the book art. My main mistake seemed to be trying to make them flat plains, trying everything from squares to nearly triangular trapazoids, none of which worked or looked right.
They match pretty close now. This was done by using the edge of the window as a guide. Extend that 3 degrees on either side (trying to match the book art again), make the edge follow a line that's tilted 45 degrees (to get the 90 degree reflection angle sited in Titan) toward the rim as well as angling it to line up with the center of spin. To get it to come to the appropriate height as seen in the canon art it has to extend to the width of the floor. Strange but true. I resisted this until I saw that it made the mirror intersect all three central cables, making my model match a feature of the book art that I thought was simply wrong: the cables being drawn as broken lines that continued at the top of the mirrors. For a long time I figured the mirrors only intersected two cables and that any art should show the outer central cable overlaying the mirror.
Doing it this way also gives the mirrors a nice bow, again matching it to the top curve illustrated in the books. The rigging cables holding the mirrors in place are 5 km thick (as always based on figures sited in the books, specifically Titan). The model now has logically consistent dimensions that make it match the book art. Weeeeeeeeeeeeee!
A last note about trying to match the book art. If you overlay my model onto the large image of
Gaea in Titan, you'll notice that the 'wheel' seems thicker than that in the art. This caused me
some consternation. I've checked and rechecked the figures. The only thing that will bring that
art to match the figures in the book is to understand that the artist forgot the 30 km thick floor (sorry Connor!).
Raise the roof in that illustration by 30 km and suddenly everything matches (at least, as well as
a CAD model and ink illustration can :)