What you're claiming is a straw man of the processes I've personally seen. In most good interviews they ask you something specific to implement, such as how do you implement a code complete feature in an IDE or an event logger. There will be some obvious naive solution that would be slow with any reasonable amount of data. They want to see if you can use the data structures in practical way.
The expectation is that if you're smart enough you can reason to a more optimal solution even if you haven't seen a trie in 15 years.
Agree...maybe I read too much into the listed algorithms, because you're suggesting real-world scenarios where some of these concepts can be applied.
But I will say, as never having used a graph search like Dijkstra's in any of my work, the idea of studying this to take a hiring test still goes with my theory of learning something by rote. If someday I need to implement the concept, or use it, I'm sure I'll go figure it out. I had to do this awhile ago with a Levenshtein distance for computing similar string matching. Had someone asked me in an interview, I'd have completely failed at it.
The expectation is that if you're smart enough you can reason to a more optimal solution even if you haven't seen a trie in 15 years.