I'm not sure if it is a poured wall or a cmu block wall at this time Jeff, but one or the other.
I've had good success modeling walls by making one pass with a simple assembly using links (width & slope) to find the top or the toe, then creating an alignment and profile along that top (or toe) and using them as targets for a more complex wall assembly.
It's true as you noted that you must manually adjust the profile (or profiles, in your case) to accurately model the tops of the benches, but it will give you fine control to determine where to end each bench.
The profile also will yield the total square footage of the wall. You can model the embedment the same way, and then get excavation quantities (I had to account for wall ex separately from the roadway ex).
In my case it was a block wall, so I had my final top-of-wall profiles step up and down to follow each course as the wall grew taller. I adjusted the profiles manually up-or-down one block height to get nice daylights that stayed within RW. The final corridor daylighted nicely, and the actual wall, when built, matched the design very well.
The one problem spot was the angle point. For the most part all the walls followed a fairly straight line except one, which had an angle point. In the field, they constructed a closure pour to fill in the gap. Will that be true in your case? If so, perhaps you do not need to model exactly what's going on at the angle points. Create your proposed profiles for the bench tops, adjust them to daylight at reasonable distances, and, extract your cut lines.
It looks good Brian. What is the secret?
First of all, miniture curves on the alignment (edit alignment geometry). Second, multiple regions with different sampling increments (inside curves have no automatic sampled lines). Third, add manual sections where they need to have additional control (frequency, add manual stations). Fourth, edit the sections manually at the angle points so they continue to have one angle point (section editor).
I was thinking about that, but did you adjust the grade of the two inside lines to something other than 2% so there isn't a bump in the grade? The new grade would depend on the angle so each would need to be calculated. Also it would cause issues if you wanted more sample lines. See attached.
I'm not sure if we are trying to achieve that level of accuracy though.
It looks like you guys are closing in with the fillets. When you get the bugs worked out can you post the design file for evaluation?