That is simply wrong, because you are making the assumption that the user will only demand their own data... But again you need all the data, and the nature of the challenge means that you can't challenge for one utxo at a time, you need to make sure the challenger pays for the reveal of the whole thing... Which will be way more expensive than putting many many many many blocks in opreturn. Also your proposal is inferior to Rootstock or BMM in sooo many ways, not the least that you have to setup BitVM with everyone who is allowed to create blocks and in BitVM3 users supposed to trust that the garbled circuits were setup correctly and the covenant committee won't cheat... That is basically a trash can on fire compared to BMM or normal MM where anyone can mine (and thus implicitly challenge DA by building on top of other blocks)... And anyone can trust that. If you think your proposal is better than 80 bytes per block, then of course you wouldn't see why putting the entire sideblocks in Bitcoin has a lot of value... You are a bit extreme here.

Actually I am glad you mentioned data on chain as a con for merge mining... That basically tells me there is no point trying to be considerate and only fees should be taken into account, there will always be someone who doesn't like even an 80 bytes extra per block fully paid for. Miners will appreciate the extra fees though if the embedded block can have enough demand to pay for it.