If they admit that their models are just that, hypothesis that can not be verified for fifty years or so. The models probably accurately model the first order effects of CO2, which appear to be nugatory at the current levels, so they add positive feedback to amplify the effect. One mechanism for positive feedback is water vapor, which most models claim will increase in the atmosphere and lead to catastrophic consequences. They apparently underestimate the role of negative feedback, for instance from increased cloud cover.
And if an inadvertent effect of CO2 burning is aerosols that cool the earth, how hard would it be to deliberately geo-engineer aerosols that would allow us to set the world thermometer just about any level we wish?
Actually not that easy, because of numerous feedback loops (as you mention) that we have no ablility to model.
You are essentially correct in everything else.
They have to assume a positive feedback loop for heating, but the evidence for that does not exist.
In essence, they make the model do what they want it to do, and call it a prediction.