You seem to be confusing nested maps (literally, a map as a value of a key) with namespaced keys, which are still just keywords, but with some magic powers. A map of namespaced keys to values is still just a single level map.
You may want to look at the official namespaced keyword destructuring example for a bit of insight into one of the ways they are commonly used - the clue is all in the name, as in they seem to be designed to group keys into semantically relevant groups, whether that corresponds to an actual namespace in your code, or is something completely ephemeral that you're inventing on the fly, up to you.
A really neat property of them is the ability to alias namespaced keywords - so, if you require a namespace, then your local alias can refer to the same keyword - ie:
(ns foo.bar
(:require [foo.baz :as baz]))
(prn ::baz/yours) ; equivalent to :foo.baz/yours
(prn ::mine) ; equivalent to :foo.bar/mine
(defn formatted-name [{:baz/keys [yours]}]
(str "hi:" yours))
(prn (formatted-name {::baz/yours "clojurian!"}))
(prn (formatted-name {:foo.bar/yours "also"})) ; same arguments, but previous isn't hardcoded.
This is definitely something you see a lot, especially in larger codebases, and helps a ton with organisation.