The point is that your questions are very basic, have been asked and answered numerous times over the years, and extremely easy to find the answer using the CPF search. How hard is it to figure out and plug in the search terms "calculate run time"?
As for driver efficiency, same deal. A simple search for "shiningbeam 1217" almost immediately tells you that it is a 7135-based driver. A follow-up CPF search for "7135 efficiency" leads you to the relevant equation of Vload/Vin.
Also if you had read more carefully, you would have also noticed that the 90 min run time is not at full power. Archer's post clearly states "the current starts to drop as the battery voltage approaches the LED forward voltage. Less lumens but more run time." And if you really did understand how to calculate run time, you'd realize that 90 min at full power with an AW18650-2600 isn't going to happen when driving an LED at 2.8A with a Vf somewhere around 3.3V. That's ~10W draw. But the AW18650 is nominally 3.7V and 2.6Ah. Obviously, 2.8A*3.3V is essentially the same magnitude as 3.7V*2.6Ah, which gives a calculated total run time of 1h at full power, assuming that Vin is higher than Vf by the necessary voltage headroom.
IMO, the way to estimate run time with a linear regulator driver like a 7135-based driver is to look at the discharge curve of the cell of interest at the relevant discharge current. Thus, you want to find discharge curves on CPF for whatever Li-ion cells you use, check the curve for say 3A discharge (assuming one is posted) and see how long the cell can hold its voltage above Vf+0.12V, which is the 7135 spec for the voltage overhead to stay in regulation. In reality, parasitic resistance in your flashlight will result in a voltage overhead greater than 0.12V, but you'll at least have a reasonable run time estimate.