Why is Depth different in OpenGL?

I get different results with the CameraMotionBlur image effect, the Camera Velocity texture seems to be generated inaccurately on OpenGL hardware (Linux). The velocity from when the camera moves to a new position seems to have a very large magnitude compared to DirectX.

From looking into it, it seems to be something to do with the prevClipPos and clipPos... but I'm not sure.

The clipPos gets Z set from the DepthTexture, and is multiplied by the _ToPrevViewProjCombined matrix to get the prevClipPos. It seems like the depth & z position coming from that _ToPrevViewProjCombined matrix are not in the same range (bias?), is it to do with OpenGL's depth buffer being -1 to 1, while DirectX is 0 to 1 ?

Any helps appreciated!

EDIT

For reference, I'm using the CameraMotionBlur .js & .shader which is a standard image effect.

I've modified it to only display the CameraVelocity Pass's output, and this is where I'm seeing differences between OpenGL and DirectX hardware (Linux vs Win)