Hokie:
This discussion got me looking at the computer program I wrote 25 years ago for retaining walls, and you are right. I also used the toe when I wrote it. That being said though, the point should be able to be taken anywhere and get the same result. The toe is just the preferred point and easier to analyze.
In looking at the OP's problem though, if the overturning moments from the soil are greater than the resisting moments of the dead load of the structure, the heel should be extended under the soil to the rear of the outfall as needed to pick up the overturning resistance of the dead weight of that soil. It could be extended beyond the sidewalls too, but not with quite as effective of a result.
And really, the soil prism is just an overall summary of the total effect of the gravity and overturning forces on the base of the structure. It is best to keep the resultant of this prism within the kern of the structure, but it can vary from there if stability and soil bearing parameters are satisfied.
Mike McCann
MMC Engineering