Great question!
First, determine the shortest path: From the boulders, under the low bar, and into the courtyard (just far enough to shoot) is approximately 10 feet. Second, determine the speed of the robot: We assumed the average speed of an FRC robot at 8 feet/second. Third, determine the time to do each task: We assumed 5 seconds (without defense) to lock on with the camera and shoot; A skilled driver, we assume, can pickup a ball with an optimized system in about 1 second.
Then the 10 second cycle time can be derived a few different ways:
1) With the rough numbers and assumptions above: Drive under the low bar at half speed (1s), drive to the boulder at half speed (1s), drive back to the low bar (1s), drive under the low bar at half speed (1s), position in the courtyard (1s), then lock on and shoot (5s) for a total of 10 seconds.
2) Do the duck walk (see Part 1: Game Theory); This method will gave us anywhere from 7 seconds to 11 seconds, depending on how fast or slow we turned. Remember that the assumption for this cycle time is a robot optimized for the task (i.e. a robot that collects boulders in the back, doesn't have to turn, drives under the low bar, and shoots from the edge of the outer works).
3) As you mentioned in your post: Look at cycle times in previous games. From watching a few Rebound Rumble (2012) matches, several teams could collect 3 to 5 balls and shoot them in 10 to 12 seconds. From a few Ultimate Ascent (2013) matches the cycle time (for a few matches I watched) was around 20 seconds; given that included filling up with more than one disc, traveling the whole length of the field, and then shooting it seems reasonable that this game's one boulder cycle time could be half of that. To get a sense for the defense in a match though, take a look at videos from Aim High (2006). It should be noted though, that by just coming off of the outer works and shooting from a longer distance, we get the protection of the outer works immunity before you shoot.
From a very non-scientific glance at a few videos (above) from 2006, 2012, and 2013 and personal experience as a driver in 2006 and coach in 2012/2013 a 10 second cycle time seems reasonable, but I encourage everyone to challenge that assumption and test for themselves once their prototypes are completed
4) If we make the assumption that at least a few elite teams will achieve the 2 boulder autonomous, then take the 15 seconds it takes to complete those two cycles in autonomous, and scale back to a reasonable cycle time; 10 seconds seems like a safe assumption.
Finally, the assumptions made in the Game Theory evaluation are just that, assumptions; that give a team a good starting point to evaluate strategy at the start of a new season. They should be revisited and challenged throughout the build phase as new data is collected (e.g. Ri3D videos, testing, scrimmages, etc).
Hopefully that helps, and if I made any glaring errors above, feel free to help me correct them!
*Edited to change assumed FRC drive speed from 10 feet/second to 8 feet/second.