I have run extensive tests with DSLR cameras, trying to approach the level of quality that you'd get from an entry level scanner. My research project in uni was on this topic. It just isn't there yet, even with these prosumer cameras and high quality lenses. The concept is good, and it works for a $30,000 hasselblad, but this will produce brutal copies from an iPhone 4S.
I'd love it if this was possible in any reasonable form as I'd abandon scanning and go to this method. In a millisecond.
Trying to justify that it is OK based on a 550 DPI is ridiculous as well - sure, the photo is large enough, but the optics inside the camera are not.
Do your own tests and see, but this is gimmicky to the extreme