It's because they would sound like crap. The glass has to be able to vibrate separately from the phone, or it's not going to be very loud. I've seen mall windows that worked that way, they sounded alright, but then they were something like 5 by 8 foot pieces of glass, and the transducer was the size of my hand.
I'm not sure if you're familiar with planar speakers - the originals were electrostatics, and later electrodynamic types were made (the most famous being the Magnepan and Maneplanars). In each case, the entire diaphram surface is driven by a magnetic field. (My electrostatics are about 4' x 1' panels fwiw, probably not much.) Those big flat panels you've seen making sound at the IMAX or over Bob Dylan's head in his concerts (he does take his own gear around) - electrostatics.
That transducer-size-of-your-hand trick is simply relying on sound propagating faster through a solid and needing a large surface area. Not new.
There are two ways to produce sound mechanically (i.e., not getting into exotics like plasma drivers) - either my pressure displacement or by amplitude displacement.
Take a sink. Place your hand halfway in perpendicular to the surface - push back and forth - pressure displacement.
Calm the water. Place a cork in the center. Bob the cork up and down - amplitude displacement.
Watch a buoy or fishing bobber sometime. It goes up and down despite the wave energy moving through it.
Air molecules are just like that. A 120 dB measured pretty close (and I know dB isn't an absolute measure and I just violated a lot there but it's what people know, so let's accept the myth) low bass sound - the air molecules are only going to move about a half millimeter - maximum. (I actually used to know the equation to calculate that off the top of my head and I don't want to go look it up).
Typical drivers most people have seen - I don't know, how about car subwoofers that you can watch pumping like pistons - pressure displacement transducers. You want loud, you need for it to be able to travel a lot.
True planar designs not trying to cheat - really do work by molecular amplitude displacement. So - virtually no movement is needed.
And yes - they do depend on size for overall frequency response (and that's true for everything, so a screen size is actually a step above anything inside a phone) - and they depend on screen size vs power to achieve volume.
I'd have to check to see what Kyocera implemented - I've forgotten.
But - it definitely was not your mall window.
Hope that clarifies what I'd said elsewhere, cheers!