I think if you actually look at research about the historical validity of science, you'll in fact find that the 90% figure is a wild overestimate. There are dozens of disciplines where virtually every single scientific consensus in the entire field has been overturned.
http://www.newyorker.com/reporting/2010/12/13/101213fa_fact_...
http://www.theatlantic.com/magazine/print/2010/11/lies-damne...
You seem to be arguing that since mankind once thought the world was flat and was wrong, and then mankind thought the world was round and was wrong (more oblate spheroidal), and then mankind thought the world was oblate spheroidal and was wrong, we definitely can't trust the scientific consensus that the world is approximately oblate spheroidal with a very slight bulge south of the equator. Indeed, maybe the earth will turn out to be cubical tomorrow, and a taurus the day after that.
"Wrong" is a relative statement. Saying the earth is flat is more wrong than saying the earth is a sphere, and saying the earth is a sphere is more wrong than saying the earth is oblate spheroidal. Yes they're all wrong, just as newtonian mechanics was, strictly speaking, "wrong", yet no one looks at a car and says "that was built using newtonian mechanics, and therefore can't possibly work".
If that were true, then the figure for scientific consensus being true would be much lower than 90%. But I don't believe that. I'd have to wonder what these scientists were doing, that I could just waltz in and blow them away with my wits and reason. I mean, what are the other scientists in this hypothetical field _doing_? Are they that incompetent or corrupt, that I can just blow a hole in their field based on casual observation? Why can't they see the same things I can?
This brings me back to that journal benchmark. Do you think you could write a summary paper discounting a substantial body of research for exhibiting various flaws? If so, do you think it would be accepted? I suggest there are few fields where it's really that easy.