Use worse case: either [(inside diameter + critical depth of cuclvert) / 2] + invert out, or actual high tide water surface.

I do usually use the worst case.  The crown of pipe or higher if the outfall is completely submerged.  But it has always occured to me that if the outfall is partially submerged that there would be less head available when the pipe is under pressure.

When I evaluate a system for outlet loss I look at the amount of submergence with respect to pipe diameter or box height.  Typically, if the submergence is minimal I do not add the outlet loss.  If you are using a spreadsheet or comp prog to calculate the hgl then you can try it both ways.  Remember that idf your system works under the worst case then you are fine.  I err on the conservative side to be safe.

You need to perform a backwater analysis which considers the level of the receiving waters.  Work backwards from the recieving waters considering their depth. The Bernouli equation is then performed in a backwards direction (upstream)after the analysis has been performed from the high point down to the receiving waters.  Use the effective coefficients of friction and head loss.  When I get back to my office I will find a reference for you to refer to.