[throwawaymaths]

The argument is about the negative space of the RE. Regardless of what the article says, To do golden gate/Gibson/etc. best practice is to cut the template first (with RE) then assemble against the open ends, so to do this you must ablate the existing re sites. The alternative is to linearize by round the horn pcr. At 33kb, it's not impossible but why bother with the pain when it's much easier to snip.