Deconstruction area layouts seem quite a bit easier to explore than construction ones, because the hardest rotations happen very early in the tape.  So I was able to find this 26 fairly quickly, mostly by studying the constrained geometry of the earliest debonding moves, placing the debonder accordingly, and then filling in with glyph and arm spots.  In that narrow domain, the biggest challenge was finding arm locations that could communicate, so that the last waste atom could be passed to the arm with disposal access.

I looked for other 27 layouts, but kept running into access problems.  In the end, I didn't find any, so I sped this one up by ~100 cycles and called it good.

I'll bet against 26 being possible, as the access problems seem too gnarly and there seems to not be enough translation directions available with only a single arm.  But I also bet folks can go quite a bit faster at 27--there's a fair amount of dead tape in my program.