With Mutable Data, the data flow itself is different. It's going through multiple hops (MaidManagers and then DataManagers), unlike Structured Data which is going straight to DataManagers. We will need to see the impacts it will have and what would be the best charging ratio.
None of the current storage costs are final in the first place (this is all pre-safecoin), we will need to work out storage costs in terms of safecoin when that gets implemented. Quite a lot is still in the design and implementation phase and needs evaluated.
The cost ratio for free mutations needs to get balanced with how much the network is getting taxed processing a mutation. Ultimately something will need to prevent spam and various approaches would have to get tried to study the impact (whether it's a one time cost that's maybe higher, whether the network gives an allocated free mutation amount over time, allow the Data object to get defined as something that has its processing cost paid for by someone else and allow free mutations by others, etc.). This of course doesn't apply to GET which would still go directly to DataManagers.
Considering the behavior of MutableData would imply safecoin has X behavior would be incorrect right now as MutableData shouldn't need to couple itself to safecoin behavior.
Safecoin or particular tag type specifications can have their own rulesets (that's what the tag type specialisation is for )