Your project consists of two separate jobs, modeling the building and compositing it into the photos.
I would model the actual building in a separate scene file in SU, and then model the site with exisiting buildings etc on sparate layers. Then I would load the other file (building) as a component into a new Layer and place it according to the site plan and exisiting builidings. Then you can easily turn on/off the various layers. No need to see the new building while matching the camera(s).
And you can easily change the building in the buildingl scene file and reload it into the site scene.
I wouldn't do anything in PS until after rendering (in LW in my case).
In PS I usually copy the background and add a black layer mask to it.
Then I load the rendered building (with alpha) inbetween that front layer and the background.
Then I can just paint on the mask to make foreground parts visible.
New renders can easily be slipped into the sandwich.
A normal, rectilinear panorama will start to look very strange, stretched and ugly once you pass around 120 degrees fov. Your panoramas are not rectilinear, and they must cover at least 160 degrees fov? You might try with cylindrical projection? Or spherical? Or even some more esoteric like Pannini? What architects usually do when drawing a street block is to draw orthographic, which isn't easy to do with a camera, and in particular not easy when you have lots of tree and other details at different distances from the camera. Take a look at some of my ortho/linear panorama photos at http://www.bknilssen.no/fasader/ . A lot of PS work was involved, and it wouldn't really be possible with your streets I think? But then you could use an orthographic camera in SU, LW or Max.
Standard procedure for shooting a normal panorama is to use a tripod with a panorama head that allows you to rotate around the lens' NoParallaxPoint (NPP). Then you could stitch tem using PTgui or some other panorama software where you can choose your projection etc.
It depends on the DWGs. If they are very cluttered with lots of small details that I don't need, and makes SU sluggish, I often clean them up inside SU. Then I group different parts and make some new layers where I put them. Next I build a "cage" from plans and elevation drawings and start modelling the building inside that cage. UV mapping and texturing is very easy inside SU, but there are no other projections than planar, and no other parameters than color. It's usually a very good start though