With the addition of the crew escape system plus nailing down the causes of their last failure, very much safer than any other manned launcher the US has had in service.
And if you want to consider the whole Saturn V STACK then you're going to have to consider counting Apollo 13 as a failure. The "payload" for Apollo is, in the end, really just the Command Module.
The whole point of doing testing with rockets is to find what you don't know, validate your assumptions... What you usually don't know is when a thing will fail - and you don't want to learn that in a real mission, so you push the test article to extreme. Another reason to do tests is something which should work - and shouldn't fail - but you can't test that without actually doing a flight test.
So, in my opinion, cherry picking test flights - where things fail routinely, otherwise why testing? - doesn't show you the designed - and updated, with test data - reliability of the system.
Regarding whole stack - I meant the whole thing which lifted off and the part of which reached low Earth orbit. That stack had the pulling safety tractor rocket. But frankly, if I'd consider Apollo-13 flight, I'd assume we have the case of mission abort in a later stage of flight, after which the existing measures were taken to bring back the craft safely. So mission was aborted - just as it would be if, say, Saturn-V grossly misbehaved on 30th second of flight - and the systems worked on safe return.
I want SpaceX to succeed as much as anyone (probably more since I work there), but one should not delude oneself into thinking that predictions are fact.