We've done it: http://threegear.com Here's a video of tracking arbitrary hand motion using a Kinect-equivalent sensor: http://youtu.be/exZ6wukQCpk
The video on your website is significantly more impressive than the linked youtube video. That being said, I still didn't see any "clicking". Perhaps I missed it.
Take a look at a typical laptop; its trackpad also can't seem to generate click events. Still people tap on it and happily ignore hardware buttons (if they're present at all). Same applies to the wildly popular touch screens. Despite the fact that the finger's projection on a touch surface is large and fuzzy, it allows for rather fine motions.
Same thing is with Leap Motion (and probably Kinect): you can define a 'click zone' on a hard surface, like your desk or screen, or tap your fingers one against another. You instantly have 8-10 'mouse buttons'. Finger touch is pretty well-defined: not only positions but velocities of the fingers change in a discernible pattern.
With a mouse, you have at least one button you can signal an event with.
Imagine doing a UI where you didn't have a mouse button. All you can do is move and point. That's a Kinect, for the most part.
I haven't used a LeapMotion, but I suspect it's the same problem; there's no way to generate a discrete event. It's all fuzzy. Did your fingers touch? Did you wave in a particular way? Some fuzzy matcher is pumping out "90% probability of event X, 75% probability of event Y" every few milliseconds, and it's up to higher layers to turn this goo into decisions that people are happy with. It's hard at all layers.
I really think you need a button, a clicker. Something "hard" in the UI that slams a voice of reason into that fuzzy tower that's continually only able to /guess/ what you're trying to do.
[We wanted a clicker on Kinect. Politically impossible. I think it would have helped a lot.]