Now my question is, is the assumption of productive hour ratio real or does it start high and end low? And if my assumption of the flat productive ratio is true, then does it actually yield any good results?
That's not the case. Hourly productivity decreases at some point.
The law of diminishing returns is at play here. Working longer hours leads to higher overall output up to a point as the marginal increase diminishes.
Then, at some point you are too tired and produces bad software that will cost to fix.
If your employer pays you a flat rate then it is profitable to make you work long hours up to the point when you start screwing up. If they pay you by the hour or if they pay you overtime, rationally they shouldn't push you beyond the point where your hourly output is too low to justify your pay.
A quick Googling for "productivity vs hours worked" led me to this:
http://bleedingheartlibertarians.com/2013/10/hours-worked-an...
But what we are talking about here specifically is the software development industry, where quantity does not always mean quality. So herein lies the problem that potentially a lot of our beliefs that "more hours worked means diminished return" could just be based on bad data analysis.