Also, are you going by age or by model year? Understand that the TCO they give is for the next five years, as if you bought that model year used now.
You could just have a car whose repair costs vary a lot.
The problem with your data, given your use case, is that you can't really compare it to anything but more of your data. I'm sure if we could see how the sausage gets made there'd be plenty to criticize on Edmunds' TCO calculator but it at least provides a way to compare different new and used cars, and it breaks down the numbers in ways that allow you to work with them a little - or compare them to experience - if you are skeptical. It's also not too hard to do a similar model for comparision for different use cases (driving only 5k miles per year brings your TCO on a reliable older car way down, for example) - you'll have your own estimates for each line item instead of their data - if you want to compare against their numbers as a benchmark.
As for Edmund's TCO vs anecdotal data from my side: I can understand your viewing it that way, but of course, I'm not going to discard almost 15 years of personal data I've collected :-) I think one of the problems I've always had with the TCO is it simply has no numbers for my use case (buying older cars but with low miles).
The other problem is that whenever I look at the TCO numbers, they seem absurd. I just looked up the numbers for my wife's newer car, which was almost new when bought, and the Edmund's numbers are bordering on absurd. We had $3500 in maintenance + repairs over 4 years (and really it's $3000, as $500 was for an extra set of tires just for snow - not really a repair or maintenance). Edmunds estimate, for that vehicle, for the same span of years is $6200 - over double the reality. I've not had $6200 of repairs in a 4 year period even for my very old, used cars. We're not stingy on repairs and maintenance - we follow the recommended schedule. In fact, her car had two expensive repairs that are probably not common for her model, and I would think the amount we spent on it is above the norm.
And she drives her car more than 15K/year.
This is why for me, having actual, real data vs some weird aggregate from a web site is more meaningful. Edmunds may be good at the depreciation amount, but I question where they get the maintenance/repairs estimate.