Hololens maps the room, yes. It does not create a "Holodeck" implementation of the room. Have you seen 3D mapping software? It doesn't image everything in 3D, it creates map data, giving coordinates to wall points, floor points, points on the couch, etc. Basically a 3D wireframe representation. It uses enough points to be able to know where all the "surfaces" are, but does not know what the actual surfaces look like. A large mirror on the wall will look just like a photo frame on the wall or an LED TV on the wall, as far as the device is concerned. It's not doing object recognition, at least not yet. It may can do certain types of object recognition from their shapes, but it is only concerned about knowing size, shape, and location of objects and surfaces in order to accurately "overlay" images.
Just want to keep the conversation going, excuse me for that, because I have huge interest in it. Object recognition is another topic, I just fantasize its potential. Questions I am trying figure out are : 1. Does the HoloLens has a 3D image of the of room? 2. What happens to object that is blocked from HoloLens sight? We know that Kinect can do 3D scan, result is a 3D model, also surface info as well (RGB image). A person would look like a himself, not just a mesh model of the person. However I also realized when I read the article on the web, the person needed to be scanned from different angles. I know it is not 3D model in strict definition, because it doesn't have the internal skeleton structure of the person, but it doesn't matter. I think the only question is whether the objects behind objects can be seen by HoloLens.
Lets say I am a developer, I am writing a game. A character has to walk along the table, but half way, it has to go through somewhere behind, say, a pensile case. Lets say there is something behind the pensile case, but right in way of the character. If HoloLens can't see those things that are blocked by the pensile case. So what I do? Basically I am facing an void, a unknown spot. If pulls the (x,y,z) at this spot, I get nothing. So best thing I can do is ignore it, keep going. If that's true, it can be an issue.
Back to the mirror thing. A mirror is not meant to a virtual mirror, it is actually taking a picture from certain viewpoint. Lets say, I am coding the Skype app, I can take picture, or live video of the room from certain angle. If my assumption is not false, it can show the room from a different viewpoint than the Kinect camera, which in front of the person wearing HoloLens. The result is like placing a mirror in front the person, with Helmet on of course, as it is supposed to be.