Just because a CISC will run multiple instructions in a single clock cycle, it doesn’t automatically make it faster. Complex instructions means complex decode logic, that makes the execution slower, even at the same clock cycle. A modern intel CPU has something like 20+ stages of pipelining, while ARM has 3-5 stages, that makes the execution more energy efficient and more powerful. Also superscalar RISC architectures exist, so RISC can also execute more instructions at a time, and in less time.

Lastly, modern x86_64 look like CISC, but are actually RISC under the hood, the single instruction is just a pseudo-instruction divided in multiple simpler instructions. I don’t believe thay makes it much more efficient.