Because the only way to get zero misconduct is to drastically reduce the amount of science that is done, probably by orders of magnitude. This is an old saw when talking about government waste: the optimal amount of waste isn't zero, because there are diminishing returns to pursuing waste, and at some point the losses you avoid by eliminating waste are swamped by the costs of eliminating it.

It doesn't follow that waste and misconduct are good, only that when we talk about policy responses to scandals, we should consider the costs involved in avoiding those scandals, and whether it's rational to pay those costs. Sometimes it is, sometimes it isn't.

The usual framing is that the optimal amount of X is the point where it is just below the cost of preventing X.

So, if it costs millions of dollars to pursue fraud, you would still be better with allowing thousands of dollars to go.