Jacob-Monod model for the lac Operon was based on experiments using two strands of bacteria which constitutively expressed $\beta$-gal: $I^{c}$(mutation in the gene lacI , which encodes the repressor) and $O^{c}$(mutation in the operator, the site where the repressor binds).
$I^{c}$ mutants are usually recessive: the Lac Repressor cannot bind to the operator, however, if a wild-copy of the gene is present in a merodiploid, the inducible pattern of expression is restored, because the Lac Repressor acts in trans (that is, it will inhibit expression of both operons when lactose is not present).
However, I've read that there's a strain ($lacI^{d}$) which is dominant, so the expression is constitutive even in the presence of the normal repressor. According to the article linked, abnormal subunits may mix with normal subunits, resulting in a disfunctional tetramer, even if a wild type lacI is also present. In terms of protein structure, why some mutations may have this effect and be dominant while other mutations do not interfere with normal copies of the protein and are recessive?