There are more interesting examples of getting your dataset wrong :
Back in the 1960's scientists sent teams to catalog animals in their environment, studying their behaviour and trying to figure out the equation for the best possible "natural balance" so that National Parks and other preserves could be managed rationally.
It went completely wrong, what looked like a perfect mix of animals went pear-shaped most of the time, despite attempts to keep the ratios they had figured out stable by adding and removing animals. In the end they realized that their assumption was wrong and that there is no natural balance, no ideal mix of x-moose, y-bears, z-gnats. Populations fluctuated and varied according to conditions. A few still cling to the idea, but the mainstream long abandoned those models in favour of better conservation methods.
Another is the uncanny ability of animals to sense their way home, especially over vast distances.
After some thorough research they found that the claims of pets having travelled hundreds or even thousands of miles to get back with their owners were greatly exaggerated. The average pet only traveled a few dozen miles at best, through semi-familiar conditions most of the time.
But the saddest part is that of all lost pets, the overwhelming majority are lost and never get even near their destination, either disappearing, presumed to have died along the way or were picked up by others. And even those who made it usually needed some help along the way with people noticing a stray animal and then notifying the owners or taking them for most of the distance to be reunited.
In fact most lost pets underperform against expectations, even if they have multiple points to base their return journey upon like familiarity, helpful people etc.
Sadly many pets are simply lost and never return. And we focus so much on those who make it back that we get a false image.