Yes, using that if() in a calculated dimension or inside of aggregation in a graph expression DOES dynamically link those two tables together for that chart only, and will avoid any connection or synthetic key on these key fields. If you hear people talking about a "date island", they're discussing doing exactly what you propose to take a disconnected set of dates and dynamically connect them to the dates in the real data set despite having different names. So it's a common solution, and may be sufficient for your needs.
But no, I don't consider it a better solution than simply modeling the connection "correctly" in the data model. It is usually a worse solution. In particular, the performance can be horrible if the tables are large. The problem, though, is figuring out how to model the connection "correctly", which can be more difficult than just throwing in an if().
If merely naming the fields the same gives you the results you want, and even if it gives you a synthetic key, I suspect that that would be the best approach. But I can't really say for sure without knowing more about the data model.
