Yes, the argument of predicting costs is both wrong and irrelevant. What is relevant is how much is spent (and agreed to spend over the next X years) and we know this exactly.
So for example, stated in the 2020-2021 annual report of the UK’s decommission authority:
our net expenditure for the year was £2,447 million compared to the original voted net expenditure of £2,606 million.
And when you look through the you see the majority of that money was not related to the UKs civilian nuclear power stations but rather its nuclear bomb making facilities.