I've had problems with this in the past too, and still come across some confusion! Here is what I have to keep reminding myself though: What matters here is not the physical shape of the eyebrow, it is the muscles underneath. Some eyebrows exist in a sharper "V" shape or soft "m" shape naturally, just like how some eyes naturally tip outwards or inwards based on genetics. This has nothing to do with type, and can certainly be confusing when typing someone based on a short clip or image.
What CT posits is that different functions end up adding muscle tone to different areas of the face based on the repeated use of certain expressions. This can absolutely get tricky to identify, because the human range of facial features is very large! Someone may look like they have heavy Ni eyelids, but it could just be that their eyelids exist within a smaller range of mobility than the general population. This is why CT signals exist in clusters, and why it's so important to check the box on every signal you see when typing. Signal mixing is possibly be more common among the P functions because they can be tricky to discern from anatomy at times, but that is okay, as long as the final count trends towards one or the other. It helps that signals will usually cluster along axes, and you can look at your end tallies to see if you have more signals along one P axis over the other.