Could you please describe your example in more details? I don’t understand your main point which is “the MMC for the original model would be -1%”. MMC will be negative only if combination of MM and your original (0 corr with MM, right?) model provides a lower COR than MM itself (so, it should be something less <-0.01). MMC will be positive, if combination of MM and your model provides a higher COR than -0.01. That has nothing in common with MM COR, it can be both positive and negative.
But in general, you must get negative MMC if your model makes MM worse and it doesn’t matter if you doing it in unique or non-unique way. Lets think about it from MM point of view. Just imagine, in near future payout will be a percentage of fund MM profit multiplied to your contribution to the MM. And someone add his own model to MM, which make a negative impact on MM (Increases loses when MM has negative COR and reduces profit when MM has positive COR). Do you think that you will be happy that this user is participating here and getting any positive payout?)
P.S. I’m not a native speaker and hope that there is nothing offensive in my style.