Sure, it’s all models. But in the case of MWI, we have two models with different mathematical behavior and you want me to label the observed phenomenon with the label from one of the models. If there were two distinct models of atoms, one that included electrons and one that had a different entity called a snark, with identical predictions from the two models, I wouldn’t have said that I knew how to observe electrons.
This may be mostly a terminological dispute, but in the original context, it was relevant for me to point out that MWI explicitly does not include wavefunction collapse and that is considered one of its attractive features.