For this project I used a 3D scan from a column in Teatro Colon (Buenos Aires, Argentina), captured with 123D Catch. The objective was to convert the scan to a pipeline ready asset and test the advantages of it's optimization. Thanks Gastón Ugarte for the scan.
In this specific case there was no significant difference when rendering:
A double or single sided mesh.
With .tex or .png files.
There was no visual difference when rendering with Catmull-Clark subdivision scheme or none.
The most efficient way to add a scanned asset to the pipeline would be to create a low poly version with the baked maps and if it is not a hero asset render it using normal and diffuse maps, as a poly with no subdiv scheme. Although rendering the high poly scan showed to be faster in the first try, having a more "friendly" geometry and lighter files is definitely a priority and a better way to optimize scanned models.