From a business perspective, presumably they would like to charge radically different amounts for only slightly different queries.
A realtor who stands to make $10,000 on a house sale might happily pay a few bucks to know the nearest schools, shops, bus stops, broadband availability etc for a property listing, but only want to do one or two searches per week.
A lawyer closing the sale of a house might happily pay $50 for a report on flood risks and nearby planned developments, if it's the most reliable data available.
At the other end of the spectrum, finding the nearest EV chargers for a vehicle navigation system? Users don't pay per search in vehicle navigation systems, even a tenth of a cent per search is too much.
Companies who are willing to talk to their customers resolve this by the sales team drawing up a different contract, and maybe even a completely different pricing scheme, for every customer. But as Google wouldn't deign to talk to a customer over a mere $100k/year they don't have this option.
From a practical perspective, structured map data often ends up with a ridiculously complicated schema, to accurately represent a reality where rules can be arbitrarily complicated. There are roads which don't allow 'trade or business vehicles except permit holders and taxis'; other roads change direction depending on the time of day; a junction might have some legal turn restrictions which don't apply to emergency vehicles, but some physical or logical turn restrictions which do. A schema complex enough to represent that sort of thing correctly will be hard to query.