As far as I know, consistency in ACID just means that the database enforces any constraints that have been defined. Since redis doesn't allow constraints to be defined, doesn't that technically make it consistent, in a vacuous sort of way?
You could also view redis's data types as a form of constraint. Using a set instead of a list achieves the same purpose as using UNIQUE in SQL.
You can still create an inconsistent state in Redis easily, even it persists well.