Given all that makes the Band a fun and interesting device...I'd like to draw an analogy...
I'm an Android guy (sorry if that's kinda rude on Windows Central). My first smart phone was the original Droid, followed by the Droid 3, and now the HTC One.
What I've noticed in the 5 or 6 years that I've owned these phones (and probably 4 or 5 Android tablets) is that while they've all been "packed" with sensors, often those sensors didn't really work. Case in point: GPS NEVER really worked on either of my Droids. I don't think this is a Motorola issue -- I think it's just a consequence of emerging technologies.
Now I just "look" at my One and it gets a GPS lock.
OK, I might be slightly exaggerating, but compared to earlier devices I owned it's rock solid. As is the accelerometer...and the camera, and touch screen, and the compass...and so on. Sensor technology for smart phones has relatively matured (as have CPU and other hardware areas), such that pretty much any decent phone on the market now "just works". Quad-core...3GB RAM...4K display...that's all overkill for the average guy!
I hear us all feverishly reverse engineering the Band, trying to figure out what the algorithm for X is, or what heuristics go into computing Y. As a human being it seems perfectly reasonable. But as a researcher, I know it's a fruitless task. There are an indeterminate number of ways to explain how any function might be computed.
Anyway, to conclude: I'm 110% sure that there are very consistent, clear, easy-to-read physiological data for detecting sleep. But the technology is pretty young, and if our reverse engineering is even partially correct, it seems that the engineers for devices like the Band are not trying very hard (which is another whole topic itself, i.e., crippling the first gen on purpose).
Give it a few more years and automatic sleep detection will be better tech than GPS.
-Matt