The numThreads option is intended to speed up restoration by performing some processing in parallel using multiple threads. The main operations that benefit are (a) restoring the data for large tables (many rows) and (b) index creation. When operating with multiple threads dependency checking is necessary to ensure that one thread is not performign an operation that may depend on an object that has not yet been created by another thread. This can slow things down in cases where there are not many large tables with many indexes.
Slow performance is most likely due to the need for a lot of dependency checking but if you file an SR with the specifics then we can look into it.