Control B: no structured data, second batch
A second unmarked page, published with this batch so each batch is compared against its own baseline.
Like the first control, this page ships no structured data at all.
Why a second one
The first control was published with the first three specimens. This batch is larger, arrives later, and enters a site that has changed in the meantime. Comparing this batch against a control from the earlier batch would mean comparing across a change in the thing being tested, which is the quiet way experiments stop meaning anything.
So each batch gets its own unmarked page, published on the same day as the rest of the batch, in the same template, with the same internal linking and the same care taken over the writing. One variable moves between a specimen and its control: the presence of JSON-LD.
What this page is about
Nothing in particular, and that is a problem worth naming. A control has to be a plausible page, because a page of filler would fail for reasons that have nothing to do with markup, and the comparison would be worthless. So this page is written to the same standard as its batch, on the one subject there is genuinely something to say about: itself.
Reading the result
If the marked specimens in this batch outperform this page, the difference is the closest this setup can get to the effect of markup on its own. If they do not, that is the finding, and it gets published exactly as readily. A lab that only reports the results it hoped for is not a lab.