You've summed up my skepticism very well. Computer models for any complex system requires endless tweaking to get right. This tweaking can only occur when the actual results can be checked against the computer model. This works great with short period systems like car crashes.
The problem with climate modeling is that the system is very long and we have no back data (ice core samples are not adequate), nor do we have a time machine to check current projections.
Simply knowing the inaccuracies of complex computer modeling (pre-tweaking), makes me role my eyes when I think about the assertions of probability given out as scientific. If you think about it, the economy is no more complex than global weather, and it too is modeled, yet you have a hard time finding two economists that agree on projections.
The inability to even closely forecast the North Atlantic hurricane season, just as it starts is all the proof I need that the computer models are either incorrect, or the inputted data is incomplete. Why anyone would put stock in the same modeling to accurately predict climate is beyond me.
I think the natural human tendency to equate effort to worth is at play. Scientists spend an enormous amount of time and brain power creating a computer model, and then believe because it is so complex, it must be valid. Additionally, as a group, they have labored for years and written thousands of papers on what effects the climate, the inclination is to self-reinforce the idea that they know enough to make solid predictions, because to admit otherwise would cast doubt on their self-worth. I'm not saying it should, just that its human to feel that way, and scientists are human.
Grad students might do that. Actual scientists have had that tendency beat out of them by their Advisors.