Working out the "cost" of anything, such as a RAIB investigation, is never going to be an exact science. Taking total costs and dividing it into the number of investigations is a lazy/simplistic average which has no practical use and shouldn't ever be used as anything more than for passing internet, and certainly never for any decision making process.
The thing is, if, say, the costs are £5m and it, say, handles 100 investigations, in that year, the lazy/simplistic average cost is £50k. That, in no way, takes into the account the difference in incidents, i.e. a Ladbroke Grove will "cost" substantially more than the South Devon rotten toilet floor in real life, so using an average of £50k for each does more harm than good.
To follow that through, in the next year, costs remain the same £5m (as most costs are staffing which are fundamentally similar year on year except for inflationary pressures), but there are more incidents, in fact, 200 incidents. The same level of staff work their socks off and manage to handle them all. Now, the average "cost" is only £25k.
Proper costing processes, allocations of costs to incidents, etc is required, and I'd hope a large public sector organisation would have proper costing accounting in place.
You really can't make decisions or evaluate performance on such a simplistic lazy basis as dividing total costs by numbers of incidents (or numbers of anything really!).
It's like the lazy "cost" of a wasted GP appointment being £50 (if I remember rightly) and the "cost" of a fire engine callout to a false alarm being £500. It's complete rubbish. I'd hope the GP would spend the 10 minutes doing something else, such as paperwork, and the "cost" of a fire engine call out from a manned station is nothing more than the diesel used!
For decision making, marginal costs are what needs to be looked at, i.e. the "extra" cost, not average cost. Sunk costs (costs that happen anyway) need to be ignored completely, and fixed costs, step costs etc need special treatment. It's variable and marginal costs that matter the most when it comes to decision making as they're the ones that are proportional to activity.