That’s not what I said at all. You can end up with a polluted vocabulary if you use your vocab with that of the reference arch. It’s easy to end up with subtle differences and false synonyms because of the inter-domain mapping that will have to happen.

Secondly, vocabulary does not help you work at higher level of abstraction. It is better to work at a higher level of abstraction if the reference architecture has a flexible meta-model. So, when you are at the meta-level you are immediately free of the reference architecture, because you can use the meta-model to create any domain architecture including the one from which you derived the meta-model. Basically, your meta-model can be the unifying thing between your domain and the reference architecture.

Thirdly, if you commit to using the reference architecture and decide to describe your intentions, it is better to create these in the language of the reference architecture. Essentially, you abandon your language.

The two key points here are (a) ubiquitous language = one vocabulary and (b) meta-model for bridging similarly close domain models is powerful because it will facilitate cross model transformations.