But the point is that Microsoft did show it to journalists during private shows after the conference. There are videos on YouTube showing how Kinect 2 works. It's not like the very first Natal video (with the family on the couch playing various games and doing stuff) where it was all just a concept.
I'm seriously surprised people are surprised that Microsoft (or any other company) wouldn't want to present those things in real time. It's a conference that is watched by millions of people all over the world. Even if you have the best and foolproof hardware and/or software (not saying that Xbone or Kinect 2 is), there're too many factors that can fuck things up that are not related with the stuff you're presenting. And the last thing you want to happen during such conference is having a giant blue screen of death on a telebim, Internet disconnecting when you want to show Skype conversations, Kinect not catching your voice properly because you're in a huge tent with hundreds of people (i.e. environment that isn't representative to the living room you will be using your Kinect in), no NBA matches when you want to show real-time updates of NBA scores in your interface etc. Even during my university years my lecturers ecnouraged us to use screencast during presentations to ommit any interferances and to make the presentation go smoothly no matter what and leave real-time demos for after the presentation.
People are bringing up the Zelda SS demo to make a point that "at least Nintendo didn't fake their show". Sure, but after that demo people on the Internet thought that Wiimote doesn't work so well with the game, that all those problems will be there when they buy a game, and that it was a big fail. But in that case neither the hardware nor the game was a problem. (IIRC) Interfances (that most likely don't appear in your average living room) caused the problem. The same with Microsoft's infmaous speech recognition in Windows Vista presentation - from what I read the problem was caused by both the ambient noise (echo) in the room and by the low-level bug with controlling the gain on the microphone that wasn't fixed at the time. While the end product (speech recognition in Vista) isn't perfect, it isn't anywhere as bad as one would think it was thanks to that presentation.