The post-hoc ROI model should be the exact same one you used in the MVPA.  This is likely why you are not seeing results.  The post-hoc ROI analysis from the MVPA results tells you the directionality of connectivity change and contributing areas that led to the original MVPA F-test results.  Try it with the [0 0 1 -1] and [-1 1] contrast weights.  This should show you what areas of the brain changed their connectivity and contributed to the MVPA obtained regions.

You could always run a MVPA with the [1 -1] and [-1 1] contrast weights, but this is a separate analysis question (e.g., group differences in connectivity from pre- to post).

Hope this helps,
Jeff