The fact that the impact is generic, and you can not be sure about that the candidate's actions were the true cause?
In this case would a more specific claim with a more solid reasoning help?
Or are you dismissing all claims of potential positive impacts?
One objection is of course that 93% is very achievable if the numbers involved are very small, but in that case the statement might just be code for "selling the office sofa" which is often seen as less impressive.
We had to re-earn their business.
In this situation I was actively evaluating support issues, customer reported problems and new feature requests to triage what was worked and when. I’m comfortable suggesting that if somebody else was making those decisions at the time, the entire company would have folded. It’s hard to expand further on that without going into a lot more detail though.
Your use of the non-statistical meaning of “accurate” as “free from error” is spot on.