Your points are valid, but I think you greatly overestimate them. Measurements are not the problem, it are the processes we still don't know about. THAT is the core of "underestimation".

But even that get's outnumbered by orders by the simple fact of how humans behave (or not).

For example I just read that somebody is producing the ozone layer destroying FCKWs again - somethign that did not happen for decades. Somewhere in east asia.
That is not predictable. (Yes, this is not climate change, but easy example.)

Also you forget that the scientists are making literally thousands of models, each with slight differences and often starting way in the past. The models that are closest to what we observe after they have run a few decades are likely to be the most exact for the future too, right?

I have been running those models on my home computers for a decade now through climateprediction.net I know a bit about that part ;)