I'm not questioning why bad conclusions exist, obviously that's money, but why would they perform an actual study and then fudge with error bars rather than, say, fudging the numbers so the things look actually how you want, or, even cleverer, fudging the experiment (perhaps even without the scientists' knowledge) so that the results look like what you want. (eg. for the HFCS experiment, to rig it simply provide HFCS as the sugar syrup, or sugar syrup as the HFCS, tada, genuine identical results!)
I really doubt that the scientists think that fudged error bars and a false conclusion are significantly more ethical than fudged numbers for the same false conclusion, and I'm pretty sure fudging numbers or fudging the experiment would be easier than fudging error bars, as well as producing a more convincing study, so why would they do it the more difficult stupider seeming way?
One possibility, of course, is "idiots", always a good answer to a "why do people do something" question. But I find it hard to imagine idiocy that endorses doing something more difficult for worse results that is also obviously more difficult and worse results. (Jokes about Microsoft Access notwithstanding.)
Then another possibility struck me, and if this is the case it's fucking amazingly brilliant and Machiavellian: if you had a study in which the numbers falsely showed HFCS and sugar to behave identically, and someone else performed a 'verification' study whose numbers differed, it would be scandalous, terrible publicity. But if you have a study with real numbers and bullshit conclusion, then any scientist who might believe otherwise comes along, looks at the study, and goes "hey, that doesn't show what the conclusion says." There's no point in him performing a study to see if the numbers show otherwise, because the numbers already show otherwise.
You know what makes news? "Hey, we did this study and it shows that other study to be completely fraudulent, and also this stuff that's in everything is terribly poisonous."
You know what doesn't make news? "Hey, the conclusion of this study from 5 years ago doesn't match with its results."
Genius. Evil, evil genius.