https://www.sciencedirect.com/science/article/abs/pii/S0165032725024346

looking at the study, basically they just found the same J-shaped correlation as found with other drug use like alcohol consumption

that is, the minority who fully abstain from alcohol have increased mortality risk and the minority who engage in excessive drinking also see increased mortality risk, but the majority who engage in moderate alcohol consumption tend to have the lowest mortality risk

The same is found here with caffeine consumption: in the dataset they looked at, most people drink 2 - 3 cups of coffee per day, and there are minorities on either side who drink much less or abstain from coffee, and a minority on the other side who drink much more than 2 - 3 cups a day.

Because most people don't suffer from mood disorders, the people who do suffer are over-represented on the margins and diluted by healthy people in the category of average consumption.

It's unlikely that actually drinking 2 - 3 cups of coffee is responsible for the positive outcomes in mood (just like we could say drinking moderate amounts of alcohol are not likely responsible for improved mortality rates), instead it's probably fair to say that the average, healthy person tends to be like everyone else and engages in socially acceptable, moderate drug use. Why they are healthy probably has less to do with their drug use and more to do with other factors like diet, exercise, economic status, access to healthcare, environmental factors (like not living in a heavily polluted place, like the way ghettos are built next to a major interstate, or rather how interstates are often built through the poorest neighborhoods), and so on.