Sound like coding to the metrics would lead to hard to read code as you find creative and convoluted ways to multiply by one and zero so to pretend you aren’t branching
Measuring complexity shouldn’t lead to finding creative ways to avoid complexity, but instead be used as a tool to encapsulate complexity well.
It could be misapplied, of course, like every other principle. For example, DRY is a big one. Just like DRY, there are cases where complexity is deserved: if nothing else, simply considering that no code used in real world context can ever be perfect, it is useful to have another measure that could hint on what to focus on in future iterations.
You are free to interpret the score within the broader context of your own experience, the problem domain your code addresses, time constraints, etc.