Why, no, deletion is possible in immutable structures, as long as nothing else references the deleted nodes. This is literally how lists / trees / any complex structures work in Haskell (and apparently in Clojure): a mutation gives you a new immutable structure, but the old immutable structure (or parts thereof) can be forgotten and disposed of.
It’s very common, however, that data will reference the data to be deleted, or be derived from it in a way that will become invalid when it’s deleted. Parallel processing becomes less straightforward because you have to make sure that all parallel processes see a consistent state of the deletion. Depending on the nature of the data to be deleted, you may actually have potential mutation, if there are keys that are sensitive data.