There are arguments for both sides. Particularly I think the correct one is without the mask because the mask is usually only a visual presentation of the die, it is not part of the die, the mask is something foreign to the die, so conceptually it would be wrong to use it. I always prefer to adopt what is conceptually correct because this tends to give less problems in the future.
Of course this is not worth if the mask is part of the data, if there is no standardization or things like that. An RG can not have mask, it varies for each state, time and other things, otherwise people will probably fill in with mask. As it is only descriptive it does not matter much.
The main argument for using it in the data is that you can simplify programming in some cases. Depending on the technology used you may have to prepare the data with the mask before using it or after receiving it.
One argument to not use is that the mask can change one day and create complicators to adapt the data.
In the past there was the argument that without the mask saves space, today this is irrelevant.