> I know rules are non monotonic, No, that's exactly not the case for Jena rules - the computation is monotonic.
We had this discussion here several times, either it was you or some other people (e.g. tina sani, kumar rohit etc.) doing the same project/exercise/homework whatever The answer is, you have to implement it by yourself in the client code - which means you have to remove the data that doesn't hold anymore. Or you always refer to only the data that will be inferred by the rules ad-hoc and don't write it back to the raw data. Indeed this might be expensive but we don't know anything about your project. This are typical design decision that YOU have to make based on YOUR requirements.
