if you say it's more an implementation problem, i have no arguments against it, since i don't know the source code for these type of mechanics ( i suppose every cluster could be associated with a variable containing the quanta types the cluster can produce, 1 type for basic clusters, 12 types for quantum towers clusters, in the future maybe dual-pillars that can produce either one type or the other...)
i am against too the solution that the harmonic produces 3 quanta for every quantum cluster because it would be overkill...
but producing one seems fair and not broken (think that if you swapped a quantum tower for a harmonic tower you could have 6 random quanta + the 3 quanta generated during the act of playing it compared to 4 random quanta )