METHODOLOGY FOR THE COMPUTATIONAL APPENDIX

Catalogue search at order 243
The script catalogue_and_overgroups.g does not assume the number of groups: it obtains N=NumberSmallGroups(243), visits every index in 1..N, and reports the result for each index. The expected value 67 is checked only at the end. The number 57 in the manuscript must not be reused as the catalogue size or retrospectively explained by an invented filter.
For each G, the script computes d(G)=log_3(|G:FrattiniSubgroup(G)|), records the count at each rank, and excludes from the subsequent search only groups with d(G) different from 2. This filter is necessary in all formulations considered, which require G itself to be among the 2-generated overgroups.
For each remaining group, it enumerates representatives of the conjugacy classes of elementary abelian subgroups H of order 9 contained in Phi(G). It is legitimate to consider H only up to conjugacy because every property tested is invariant under conjugation in G. When H is required to be central, each conjugacy class is a singleton.
For each H, however, the script enumerates all actual subgroups K of G, expanding each class returned by ConjugacyClassesSubgroups using Elements. It selects the K satisfying H<=K and d(K)=2 and checks whether these are exactly H and G. Printing only representatives of the conjugacy classes of K would not suffice to establish the actual number of overgroups.
The search reports three properties separately: (i) H=Z(G) is isomorphic to C3xC3 and is contained in Phi(G), as in the original theorem; (ii) H is isomorphic to C3xC3 and is contained in the intersection of Z(G) and Phi(G), the order-243 instance of the corrected central-subgroup formulation, whose general statement also accommodates p=2; (iii) H is isomorphic to C3xC3 and is contained in Phi(G), without a centrality assumption. For every H that fails the property, the output records the orders and multiplicities of its overgroups and an intermediate 2-generated overgroup K witnessing the failure.
Any exclusion must refer to the precise property tested. No finite search at order 243 establishes nonexistence for all groups of larger order or for all 3-groups.

Reconstruction from the presentations
The presentation generators are constructed in a free group, using the GAP commutator convention Comm(u,v)=u^-1*v^-1*u*v. The presented group is converted first to a permutation group and then to a pc group; the named generators are transported through both isomorphisms. The subgroup H is defined using these transported generators, rather than arbitrarily chosen generators of a SmallGroup representative. IdGroup is used as a subsequent check, not to identify the presentation letters with the catalogue generators G.1,G.2,... without verification.
For the p=2 and order-729 examples, the script checks |G|, d(G), Z(G), Phi(G), the order and invariants of H, H<=Z(G), H<=Phi(G), and whether the actual set of 2-generated overgroups is exactly {H,G}.

Orbits for A5
The script a5_orbits.g enumerates all ordered pairs (a,b) in A5xA5, retains exactly those satisfying |<a,b>|=60, and computes their orbits under simultaneous conjugation by S5. It also checks |Aut(A5)|=120. An automorphism fixing a generating pair is the identity, so each orbit has 120 elements. The output includes representatives of all 19 orbits and their distribution according to the orders of the two entries. The independent Python verification yields 2280 pairs, 19 orbits of size 120, and the matrix [[0,1,2],[1,1,4],[2,4,4]]. This numerical verification concerns A5 and does not establish an extension to all nonabelian simple groups.

GAP COMMANDS
Run from the scripts directory, with GAP and SmallGrp installed:
gap -q -b -A -T catalogue_and_overgroups.g > catalogue_and_overgroups.out
gap -q -b -A -T a5_orbits.g > a5_orbits.out
-A disables package autoloading; the first script explicitly loads SmallGrp. Both scripts print the GAP version, and the first also prints the SmallGrp version. The computations are deterministic and use no random seed.

REPRESENTATIVE CHECK AT p=5
The script sample_p5.g performs a representative check at p=5. Since H=Z(G) is normal, it enumerates all subgroups of G/H (of order 125) and takes their preimages. The correspondence theorem guarantees that these are precisely all overgroups of H. There are 39 such overgroups, of which only H and G have rank 2. The symbolic proof, not the representative check alone, establishes the result for every p>=5.

WITNESSES FOR THE EXCLUSION
For each catalogue candidate H that fails, the script produces a pair of elements generating an intermediate subgroup K. The pair is chosen deterministically: the first u outside Phi(K), followed by the first v outside <u,Phi(K)>, in the order given by Elements(K). By the Burnside basis theorem, u and v generate K; the script also explicitly checks Group(u,v)=K.
