Sorry, I've just re-read it and it also reads like garbage to me
What I meant to say was to use the real video in Lightwave at a lower resolution to actually place your scene objects, simply rendering the 3D parts of the scene (such as the objects, shadows their associated lighting). This should give you a very plain 3D output (just make sure you render to a 32 bit format to keep the alpha channels). Then, recombine that with the original video stream in a separate compositor (after effects, DFX, Combustion, etc...)
Generally this will give you more control and consequently a better quality output.
If, on the other hand, you don't have access to a compositing package, then just do it in Lightwave....