Confusing failure modes with effects and causes in FMEAs

[McLeod]

FMEA resources online indicate that a common mistake is to confuse the failure mode with the effect or the cause. I've seen this happen quite a bit, and it's usually due to the fact that the cause-and-effect chain is longer than three discrete items for the failure mode in question. So what is the best practice with respect to capturing the "right" failure mode and the necessary levels of cause and effect on either side of it?