First off, none of us AFNI drones have ever seen or tried to analyze data such as you describe, so we don't really have any idea AT ALL if these registration steps will work well enough for your purposes. You need to somehow establish a test-bed system where you know there is "activation" and then see how well you can find it with this approach. In the early days of FMRI, the test-bed systems were motor cortex (finger tapping experiments) and visual cortex (flashing light experiments). We already knew where the primary motor and visual cortexes are, so we could look for signal changes there after processing the data.
Second off, 3dAllineate -nwarp poly9 is experimental, and is quite likely to run amok and mangle your images. You need to carefully examine the output to make sure that the registration worked well. Which is what the last lines of output from 3dAllineate always say! I put that caution there because it is my experience that people assume that software always gives right results, and then bad things happen to them.