tldr: less is more; in general, severely limit landmarks to keep the maps useful.
In my mind, a landmark is generally something that was there 50 years ago and is likely to be there 50 years from now. It is something that is both a significant destination for a majority of the population likely coming to an area as well as a significant visual presence that lends itself to navigating in an unfamiliar area. However, landmarks should be few-and-far-between so that when a user is looking at a map while driving, they aren't forced to process clutter in order to discern their route (this is especially important in denser urban areas).
As such, I'd suggest tempering many of the items in the proposal with a minimum dimension requirement, perhaps modulated with the local population/building/road density as well as with the local building size (where the latter would require that the landmark be significantly larger than surrounding structures) and with the number of existing landmarks in the vicinity.
The first part (minimum dimensions) would help to curb a recent local surge in the addition of "pocket parks" -- mini-greenspaces that, while truly being public parks, are so small as to be indistinguishable from a home or business lawn. I've been toying with a guideline that goes something like "In general, only parks that have actual roads/driveways/parking lots should be mapped"
The first part would also eliminate the "university branches", DMVs, urgent cares, etc. that are little more than a recently vacant storefront in a stripmall -- all of which should be easily located using the in-app search.
The second part (population/road/building density) would keep someone from mapping every DMV, police department, fire department, public restroom, etc. -- again, arguably all things that should be located using the in-app search. (NOTE: the argument that someone would use waze to locate a hospital/fire/police department in an emergency is difficult for me to accept-- in an emergency, I would hope that a user would call 911, not try to search around on waze -- which could be horribly out of date or even devoid of information because of data-service outages -- to find the closest first-responder outpost; I won't even speculate about the liability issues there).
The third part (number of local landmarks in a given area) would also help prevent visual clutter in state capitals, county seats, major metro areas, etc. where there can be a major concentration of public buildings, museums, sporting venues, etc. -- and where making every one of them a landmark means that you end up with a map of landmarks that has an occasional street peering through vs. the more useful-for-navigation map of streets with an occasional landmark.