A rational thought may not be a predictive thought.
Before people knew the earth was a globe, it was entirely rational to think everything that goes up goes down again.
And for all testing you can do, ie, throwing an apple up, it'll come down again.
It is entirely rational to conclude from the available evidence that everything falls down when thrown.
But recently we've discovered that things you throw up hard enough may not fall down at all.
Thusly, the rational thought was still rational but wrong.
Predictability is important when you want to evaluate scientific models (ie, assume earth is flat, assume earth is round, assume earth is a mostly round potato all have different predictive quality, though the first ones are easier and sometimes sufficient)