Scientific fraud doesn't make sense to me

Having just read about an alleged data manipulation scandal in political science in the USA, I wonder once more why people are committing scientific fraud. I get plagiarism in the humanities, although of course the risk of getting caught is growing ever greater considering the ease with which text searches can be conducted these days. Still, I realise why people might be tempted into doing it, especially if they are not pursuing a career in scholarship, just want a few extra letters to appear on their business card, and are unwilling or unable to make the necessary effort. Inventing or manipulating empirical data as a career scientist, however, I don't really understand.

To me there seem to be four factors involved in somebody's decision whether to commit fraud: (1) ethics, (2) the rewards of getting away with it, (3) the likelihood of getting caught, and (4) the consequences of getting caught. Honest people will, of course, not commit fraud, so let's limit our considerations to the dishonest minority. Likely there are people out there in pretty much every profession who will behave badly if they think they can get away with it and if the rewards are sufficiently high, even if they are vastly outnumbered by honest colleagues.

That leaves factors 2, 3 and 4. I will now make another simplifying assumption, but one that I consider to be very realistic: if somebody is found to have fabricated data, their career in science is pretty much over. They can forget about getting serious grants or about being hired by a serious institution. In other words, the consequences of getting caught are always catastrophic. Factor 4 is thus invariant across all possible scenarios, and what really matters are factors 2 and 3, the reward of not getting caught and the likelihood of getting caught. What we are searching for is a scenario in which the former is high and the latter is low, because that would provide the incentive to commit fraud.

The thing is now that at least to me these two factors appear to be strongly positively correlated. On the one side, somebody might fabricate a spectacular result in an extremely important research field, such as "newly discovered substance cures cancer". That would make the rewards of getting away with it high - fame, big research grants, perhaps a high profile job. But on the other hand, these being spectacular results in an important research field, many other people will by definition try to reproduce the results, they will try to build on them. And then the whole house of cards comes crashing down. High potential reward means high risk, and in fact I would say near certainty, of discovery.

The alternative is to fabricate unspectacular results that people are unlikely to try and replicate, e.g. a technical report on the structure of a new substance that has no known uses. But then, what is the point of making stuff up if nobody is going to care? Low risk means low potential reward, but that also means that even a low likelihood of catastrophic consequences looms large over the minuscule benefits of cheating.

And this is what I don't get. Even assuming that somebody is crooked and has no qualms fabricating data, what do they hope to get out of it? Why do they think they can get away with it? If there is any significant advantage to what they are doing compared with doing honest science they will be caught sooner or later, and then they will fall much deeper than they have ever risen.

And the weirdest thing is, to even get away with it in the short term they must be highly proficient at scientific writing and at understanding study designs, because otherwise they could not make up convincing-looking results and publish them. So why not just do honest science if they have that capability?

Or am I just incapable of figuring out their true motivation? Perhaps everybody who does that just hopes that nobody will catch them...