The S thing included more than just translations. Rotations, dialations, and combinations of the three were considered.
S can't have uncountably many of itself in the euclidean plane unless it's defined as any of a certain shape constructed according to certain rules but with variable parameters. If those parameters vary in the right way, uncountably many S's are possible.
I had arbitrarily chosen two 3/4-circular arcs, conjoined only at one pair of endpoints to be C¹. (I did notice that two semicircles joined in the same way can be translated by ε>0 without self-intersection.)