Moore's Law is called a law the same way many other laws are, such as Power law, Pareto law, or Swanson law. It is an empirical observation that some processes behave in a certain way. I agree it is not the best of terms to use, but it is what it is.

I do want to highlight, though, that while transistors are indeed reaching a size limit, this is not substantial enough to believe that Moore's law is fading away. The key observation is that Moore's law is really about doubling computing power, not transistor counts per se.

Once upon a time, the main force driving computing power growth was clock speed. Then we reached a limit around 4ghertz, then we started expanding to multi-core architectures, then to out-of-order execution. Currently, we might be undergoing a shift back towards RISC architectures. We also greatly improved our memory and storage devices, which are the performance limitation of most applications. There is also research on stacking transistors, creating 3D grids that could easily double or triple current standards.

So, indeed, it is getting harder to make smaller transistors, but who cares, really? That's not the only option we have to improve performance