I think there are two points to be made here.
(1) Any machine generated index stored electronically won’t have any need to use cross references. Users will enter a word or name or something and the software will return all the use instances, or perhaps jump to the first (most likely?) target. In the case of green beans you won’t be able to look for beans, green because the index generator will never have encountered that phrase. You could still enter beans, green, but you would get a list of beans and a list of green.
(2) A thorough hand generated index is an order of magnitude better than a machine index. Just look at an index in almost any software user guide. Cross references? What are those? We humans can grasp that green beans might also need a beans, green entry.
With that said I prefer an electronic “index” because it’s a lot faster than a manual lookup.
= = = = =
A long time ago I assisted the publications department for six months. What did I do? Hand generated indexes for the manuals. Tedious work but at least I could use a search capability to find the use instances for the index entries. Best of all was that I decided what would go in the index, and I decided a lot of things needed to be in there.