How do I catch the cases where the stats mislead me?
You can model most problems well now. What still catches you are the edge cases, the moments when the numbers look right but quietly lie. That worry is a good sign, not a bad one. The people who get burned by misleading stats are usually the ones who stopped wondering whether they might be wrong. You have not stopped, and that instinct is worth trusting. Statistics mislead most often when the assumptions underneath a method do not match the data in front of you, and nothing on the surface warns you. The result looks clean. The chart looks convincing. The trap is in the fit between tool and situation, and it is invisible unless you go looking for it on purpose. So build the looking into your habit. For each analysis, write down what would have to be true for this method to be trustworthy, then check whether it actually is. Ask what a skeptical colleague would poke at. Try the result a second way and see if it holds. You will not catch every trap, but a mind that keeps asking where am I being fooled catches far more than one that assumes it is safe.