This isn't actually a problem unique to conlanging. Linguists working with natural languages run into this all the time too! Fundamentally, the goal of your terminology—whatever you end up using—should be to give the audience a clear understanding of what you're talking about.
If your cases happen to work very similarly to certain cases in Latin, then using the traditional Latin names might serve this purpose. It means people who already speak Latin (or Greek or German or…) don't need to learn a new name for "this is the case for the subject of a verb". When that happens, using a term like "nominative" can be helpful. There are a whole lot of languages in the world that have a "nominative" case, so most linguists are familiar with it.
However, most cases don't work quite like certain cases in Latin. When this happens, you have to decide if it's worth using a name that's close enough, even if it's a bit misleading, or if it's worth creating a new name that your audience needs to learn. Which way this tradeoff swings tends to depend on how close your thing is to the traditional meaning, and how common the established terminology is. (A lot more people know "dative" than "adessive".)
Personally, I would err on the side of using more accurate names for things, even if you need to create new terms for it. When two languages both have a "dative" case because it's used for the recipient of giving, even though in one language it's fundamentally a case about beneficiaries and in the other language it's fundamentally a case about circumstances, the standardized name is actively hindering readers' understanding. Calling them the benefactive and circumstantial cases could avoid that.