Following the problem with mirrored blend shapes I repeated the work yesterday recreating the eyes as a pair and manipulating their clusters individually to try and create as close to symmetrical expressions as possible. The resulting video is not so bad. But I wonder if there is an easier way because it means each eye must be key framed separately. And when keying via blend shapes the key frames cannot be seen on the timeline.
Currently I have not pulled the eye shapes around so much when they pass through the face as they appear to witht eh 'surprised' eyes. Partially because I have yet to animate the eyebrows and therefore the brow of the head, but also the cheek bones. This will hopefully lead to a better fit. And if not then I can tweak the blend shapes once more to match.