Logging the last ten jobs as estimate to actual ratios is a simple habit that most teams never start, and the numbers make the case better than any argument. A median of 1.225 with ninety percent of jobs overrunning is not bad luck. It is a bias you can correct.
The point that the gap between P50 and P80 is the contingency on fixed price work is the part I would underline. It turns a guess into a number you can explain to a client. Separating change requests from estimating error matters just as much, because mixing them hides both problems.
On fixed price client work we pair this with tighter scope. Each piece is one defined problem with written acceptance criteria, so a change request is easy to see and price on its own. That keeps the history clean enough for the ratios to mean something.
Do you track the ratio per type of job, or does ten jobs become too small a sample once you split them?