AtlThrow inside a COM method implementation

Considering code something like this (the implementation of some COM
interface method using ATL):

<code>

HRESULT CSomeClass::DoSomething( /*[in,out]*/ long* count,
                      /*[out]*/ SAFEARRAY** result)
{
    if (count == NULL || result == NULL)
    {
        return E_POINTER;
    }

    CComSafeArrayBound bounds[2] =
    {
        ...
    };

    CComSafeArray<VARIANT> data(bounds, _countof(bounds));
    ...
    ...

    *result = data.Detach();
    ...
    return S_OK;
}

</code>

I think (but I'm not sure) that there is a problem in the use of
CComSafeArray constructor above.
In fact, if some error occurrs during constructor, what does happen?
Spelunking in CComSafeArray constructor overload code, I read in
<atlsafe.h>:

<code>

    // create SAFEARRAY where number of elements = ulCount
    explicit CComSafeArray(ULONG ulCount, LONG lLBound = 0) : m_psa(NULL)
    {
        CComSafeArrayBound bound(ulCount, lLBound);
        HRESULT hRes = Create(&bound);
        if (FAILED(hRes))
            AtlThrow(hRes);
    }

</code>

So, in case of errors, AtlThrow() is called.
Reading the documentation, if my understanding is correct, AtlThrow throws a
CAtlException in ATL projects.

So, should the original code posted above be modified, wrapping the
CComSafeArray constructor call in a try/catch( CAtlException & ) block?
e.g.

<code>

   HRESULT hr = S_OK;
   try
   {
     ... <above code> ...
   }
   catch (CAtlException & e)
   {
      hr = (HRESULT)e;
   }

   return hr;

</code>

Or does ATL have some C++ template "magic" to automatically return an
HRESULT value when AtlThrow is called?

Thanks in advance,
Giovanni