singleton in a DLL loaded class

I have a class loaded from a DLL that I would like to implement as a
multi-threadable singleton. Unfortunately, I seem to still be getting
two distinct instances of this class. It is implemented as below.

The only funny thing I am doing is to load the DLL from two threads
simultaneously (thus the need for the singleton). But, according to
other posts I've read, the OS will point them both to the same
instance of the DLL so that only one thread actually loads the DLL.

Can anybody point out where I may be going wrong?

Thanks,
PaulH

//singletonclass.h
class SingletonClass : BaseClass
{
public:
    static SingletonClass& Instance();
    ~SingletonClass();
    //... Implement BaseClass functions
private:
    static std::auto_ptr< SingletonClass > _theInstance;
    static Mutex* _m;
}

//singletonclass.cpp
std::auto_ptr< SingletonClass > SingletonClass::_theInstance;
Mutex* SingletonClass::_m;

/*static*/ SingletonClass& SingletonClass::Instance()
{
    Mutex::Create( &_m );
    _m->lock();
    if( _theInstance.get() == NULL )
         _theInstance.reset( new SingletonClass() );
    _m->unlock();
    return *_theInstance;
}

//dll.cpp
MY_API BaseClass* Create()
{
    return &SingletonClass::Instance();
}