Yeah, BP II is a very tough call if it is hard to get a clear history, and I think in the absence of collateral it is a matter of evaluating other possible diagnoses and finding them wanting. This is definitely not ideal but it is what you are left with. Zimmerman et al. (2019) suggests two questions about affective instability that have a very high negative predictive value for BPD in people with mood disorders, but the specificity is garbage, so it only does a bit of work for you.
It is obviously bad from a medication perspective to miss a BP I diagnosis, but the data on switch rates wrt antidepressants in BP II is not that compelling, so it is a question of how depressed they are, are they responding to antidepressants, bad past rxns to antidepressants, their beliefs about antidepressants and their diangosis, etc. I like working with this population but I always feel the need to very carefully document my treatment rationales because I live in fear of passing them along to a less careful or thoughtful resident who will say "oh this borderline doesn't need all these medications, let's take them off" with predictable results.
At the end of the day if you can't get good history and you can't get good collateral I think you may be stuck with eliminating alternatives and attempting Jasperian verstehen, which is not especially satisfying from a psychometric (or insurance) perspective I'm sure.