This is an intriguing article from a generally reliable source, Psychiatry Weekly. The fact that it is plausible because it ties in with other things, is from a good source, and would be really cool if it proves out makes it a good example of why we exercise caution with studies, even if they are moderately robust and tell us what we want to hear.
Even NEJM and JAMA studies can turn out to lead nowhere.
When effects show up in one part of a study but not others, it might mean that some interesting subtlety has been identified. More likely, it means the entire effect is weak or nonexistent.
Even if true, and destined to lead us to new understandings of a subject, statistically significant effects may not be the most important effects we should be putting our effort into.
Follow the footnotes to see their weaknesses. In this case, #2 also has an effect that shows up in part of the study but disappears under other conditions.
In this case, overgeneralised memory, which they define rather uh, overgenerally in the study, may be related, even mildly predictive of adolescent depression. It makes some sense. Moderate depression* interferes with recall and storage - abuse and trauma do also - that the brain would make some memories less retrievable seems possible.
*Interestingly, mild depression does not, and may even slightly improve memory. Grouchy people really do see the world more clearly. However, euthymia or even hypomania, even when not realistic, seems to be adaptive anyway. The gains from the successes of being unrealistically positive seem to outweigh the gains one gets from conserving effort on lost causes. But that's more sociobiology than psychology.