This would allow doing something like

// in Thread A:
int lockDepth = myMutex.unlockRecursive();
mySemaphore.acquire(); // Wait for thread B to access threadA and to release the semaphore
myMutex.lockRecursive(lockDepth);

Since the thread A method called by thread B will lock and unlock the mutex as well it is required to completely unlock the mutex first, which right now is difficult to implement.