> As of 17 September 2016, there have been 310 manned spaceflights [...] 8 of which were sub-orbital [1]
There have been 302 manned orbital launches according to [1] and from adding up the successful launches in [2] there seem to have been a total of 5057 orbital launches so far, in the 70 or so years of spaceflight.
Apparently the Mars fleet needs from 3 to 5 tanker launches per manned craft, so this program will require between 4000 and 6000 orbital launches, so around 5000 - which is about the same number as the total number of (successful) human rocket launches to date!
Looking at the table in [2] we can see that the best success rates for manned launch systems are 97% for both Soyuz and the Space Shuttle. using this as a worst-case estimate, out of those 5000 launches we can expect 150 to fail. Assuming one in five are crewed, with 100 crew per vehicle, that means 3000 deaths over the life of the program.
I guess that we can expect improvements in reliability with a program this size, and using a mass-produced vehicle which is re-usable is not something we have safety data for in the space-flight arena yet. You would hopefully expect reliability and safety to improve, getting towards the standards of modern aircraft, or at least experimental aircraft or military systems perhaps. But even three nines (99.9%, or one failure per 1000 lanches) is still multiple orders of magnitude better than we can achieve with spaceflight right now, so SpaceX are setting a pretty tough goal for themselves here. Still, I really do hope they succeed, and it will be exciting to watch them try...
Launch abort should provide a safety margin on launch. It does less well in other segments of the mission, such as deep space, reentry, or landing. If you're looking at deep space missions, there've been issues at all of these.
There are also the instances of missions (Japan fairly recently in a Venus probe) where issues affecting an unmanned flight but not causing failure of the mission, would result in failure of a manned flight (expenditure of all life support resources).
The estimates here do suggest some significant loss of life will be expected. That's not unheard of in voyages of discovery, colonisation, and exploration.
