Meanwhile the assertion that models fit past events is near irrelevant since that is data which is already known and it is expected that the models would have been adjusted in the first place to fit that data). For example, I can construct an interpolation of any temperature (or other numerical) data to perfect precision using an even degree polynomial of sufficiently high degree, yet it'll be completely irrelevant once I attempt any sort of extrapolation into the future (odds are good, about 50% I'd say, that it'll predict temperatures far below absolute zero by 2100).
Shockingly, scientists are aware of that issue, and have developed methods to test models against existing data. They do that by training on one chunk of the available data, and testing against another.
You're making two more mistakes in your analysis.
One, you complain that models that fit old data perfectly are wrong because all they do is fit data. Then you complain that the models don't fit the data perfectly - precisely because they don't just fit data. Which is it? You can't have it both ways.
Two, you think that we have direct measurements for everything. We don't. We'd like to, but we don't. And even the direct measurements we have need to be transformed into data that can be compared across measurements. All of that is subject to being wrong.
This profound inability to admit error is why I don't trust current climate models or the doomsday predictions they spawn in the least. That's why I'm going to wait a few decades and see what happens. If it genuinely is as bad as claimed, then we'll see something by then.
Unfortunately, that inability to admit error is only in your head. The models have been changed countless times over the last decades, and have gotten better in response. Lastly, if you wait a few decades, it'll be too late to head off any meaningful changes. As the joke goes: what if we'd make changes for a better planet when it's not necessary?