Fast Screen Capture and Lost Vsync

I would like to create a real-time movie with a homemade application that performs quick screen captures using the portion of the screen occupied by the running 3D application.

I know there are already several applications for this (like FRAPS or Taksi) and even dedicated DirectShow filters (like UScreenCapture), but I really need to do this with my own external application.

When properly configured (UScreenCapture + ffdshow), capturing full screen compression does not consume as much CPU as you would expect (around 15%) and does not degrade the performance of a 3D application.

The problem with doing capture from an external application is that the 3D application loses its Vsync and creates a shaggy, difficult to use 3D application (the 3D application is only present on a small part of the screen, the rest is GDI, DirectX)

FRAPS solves this problem by allowing you to capture only one application at a time (with focus). Depending on the technology used (OpenGl, DirectX, GDI), it grabs Vsync and captures it (with glReadPixels, ...) without breaking it. This does not solve my problem as I want a fully composited screen image (including 3D and others) and a smooth 3D application.

UScreenCapture seems to use the DirectX shortcut to capture the entire screen, but the openGL 3D app is still out of sync.

BitBlt execution is too slow and CPU is too slow to achieve realtime 30fps (at least under Windows XP, not sure with 7)

My question is to know if there is a way to achieve my goal with Windows 7 and the new DirectX layout engine?

Windows 7 will manage to show live VSynced duplicate previews for each application (in the taskbar), so there must be a way to access the displayed currenlty screen buffer without breaking the rendering of the OpenGL 3D application?

Any other suggestion, technology? thanks