When GR-F-CLZ-L has an offset from EP-L locked at -20' (as shown in both screenshots above, for testing), this template processes as described above: the whole side slope is unbroken at 1:6, but at least some solution is found. The same is true when the offset is locked at -10'.
When GR-F-CLZ-L has an offset from EP-L locked at -30' (as the typical section calls for), this template doesn't process the fill condition properly and falls back onto that bogus condition that Rock-L is part of.
There are no other distinctions between (partial) success and complete failure.
It's clear to see that adding 10 feet to that 14' dimension won't cause any distinctly different solution, whether the further slope is 1:6 or 1:3, so why does it vanish?
I thought this might break the solution, but if the offset is locked at -10' as above, and I lock the vertical of GR-F-TIE-L to +1 from ES-L--meaning above the shoulder point--the 1:6 slope persists.