In a new edition of Funcoids and Reloids article (section “Some counter-examples”) I wrote a counter-example against this conjecture, upholding that there exists a reloid with atomic domain, which is neither injective nor constant.
The conjecture is equivalent to this my MathOverflow question, which was quickly solved by my colleagues. I just adapted the proof for my conjecture.