Correct. I think reversion to the mean is a non-sequitur here.

Neither the GP's quote, nor the OP, are making the statement "I understand that this mistake was an outlier, you'll probably never make this mistake again if you remain unchanged".

The claim being made is that acknowledging the mistake and learning from it can dramatically reduce the base probability of such mistakes happening again.

Regression to the mean applies any time a process has high variance. If I have an unusually productive week, odds are the next week will be worse, simply because the previous one was an outlier. But people often make the mistake of noticing week X+1 was much worse than week X, and attributing it to some variable that changed between X and X+1.

In my model, humans have an average performance, that increases over time (unless something terrible happens) and a random component, that makes performance fluctuate around the average.

The question is: is the day-to-day random variation much bigger than the day-to-day average increase? If so, regression towards the mean makes total sense.