Shadow catcher is just shadow catcher, not matte object. The fact Vray combined shadow catcher and matte object in one solution was the reason it was so confusing to use, and that it did not work correctly up until 3.00.07 version.
More can be found here:
http://forums.chaosgroup.com/showthread.php?78565-Correct-workflow-if-CG-object-integration-onto-backplate&highlight=correct+workflowShadow catcher is shadow catcher, and is already implemented. Matte object (with possibility to catch shadows) is a different thing, that is not yet implemented. Simple as that.
If you combine these two together, you get as confusing solution as Vray has, which needs a lot of controls (Vlado is going to add two more buttons in new version just to make it working right)
That being said, if you use shadow catcher properly, as it's supposed to be used, then you should lose as little info as possible.
So if you have already rendered your interior, and want to add CG character into it, then:
1, Place spherical cam into center of your interior and render out HDRI. Doesnt have to be large one, even 3k*1.5k will do.
2, Create one shadow catcher material, and use your final image as screen projected backplate
3, Assign this shadowcatcher material to every object in your scene. If you haven't moved your camera, it should match perfectly.
4, Add HDRI as your environment map, and rotate it to match the orientation of the room. Keep in mind HDRI will not be used only for stuff behind window and such, but also for everything out of camera frustrum, where there is no backplate information.
5, Optionally, if you have reflection render element, you can edit it (make it monochromatic) and then use it as reflection map for your shadow catcher, so your CG character will reflect in the rest of your scene.
If you do these things right, then you should lose as little information as possible, and your character will really look like it correctly belongs in the scene, receiving all the correct GI bounce light and such. Sure it's a bigger pain in the ass than just selecting everything and clicking matte checkbox, but you are also rewarded with a lot faster render time. Think of it as a workaround before matte objects get implemented.
But mainly, when you separate stuff, you will always break something. Even Vray's matte solution isn't perfect because there are huge problems when compositing colored shadows. When you render everything in one piece, then you always get the best, most accurate results. Then if client want's changes, which do not touch geometry, then you can use render selected feature (render subset in Mental Ray, render mask in Vray) which will soon be in Corona as well.