I think the original HEC2 intent here was to allow some kind of recognition of the variation in distance between sections in bend locations. The outside bend distance being greater than the inside bend distance and the channel distance. All these are composited in determning the loss between sections and a single WS elevation at the next section.
The willy-nilly result has usually been setting the bank stations at the top-of-bank location, however as far as the program goes you can set these locations anywhere that is meaninful for your investigation. If you are only looking at bank full flows and less, you might want to divide the channel itself into 3 parts to study shear/power distribution on the banks.
Ron Kilmartin
ronkilmartin@attbi.com