We don't need them, but sometimes they aren't avoidable:
An example is a counter (e.g. a Johnson counter) used as clock divider. If not all possible states of the counter are used (e.g. a decade divider made from a 4bit counter, which uses only 10 of 16 possible states), it is possible that this counter might enter into an unwanted or
false state (e.g. by an SEU or an ESD event), from which it (possibly) can't bail out if not enforced by design.
Such a circuit clearly has one or more
false paths, which will be detected by the fault coverage tool, because they really exist. This information should warn the designer that he possibly should worry about and see to arrange for a fail-safe solution (if the circuit can't resolve it from its own). Not fully used Johnson counters actually might need additional (reset) hardware to help them out from a false count cycle.
Generally this applies to all circuits which aren't "fully used", i.e. which include actually -- for this dedicated design purpose -- not needed, but possibly accessible
false paths.