Mocking a six ton payload is actually fairly expensive, to get the cg, vibration etc. right. A lot more difficult than payload = mock(satellite); NASA did it for the Hubble but so the ground crew could practice operations and not drop the billion dollar space observatory, not to try and save some money if something blew up during testing.

Surely still much cheaper than a two hundred million dollar satellite?

And at what frequency would it save the payload? In 85-95% of cases, payload is safely delivered to space, so the dummy just slows things down. In most of the remaining cases the payload is lost on it's actual ride into space (e.g. the last time SpaceX had a problem, CRS-7) where a dummy can't help. There have been a handful of pad accidents involving loss of payload, but they are very much the exception to the rule about the exception to the rule.

It is possible to damage the rocket and systems when mating the payload in the faring, which would not become apparent without further testing and somewhat negates the purpose of this test itself.

The payload is mated early to save time by speeding up the launch schedule [1] so a mock payload would have the opposite effect - it would need mated and then removed after the test fire.

[1] https://twitter.com/pbdes/status/771411924907094016

The question nobody is answering....