That is true. The basic rule is, the network you are connected to is connected to you and knows something. What @reC suggested - altering signal strength of the modem - sounds like it could be used to cause some ambiguity there. But cellphone system (and location info based on it) is probably too much of a challenge because its design requires known connection.
But cell connection is only one culprit, so why should it be used as a reason not to also take into account the others: GPS/GNSS, wi-fi, BT and even gyroscope (movement)? Since we can kill cell modem in L5, we’d be in a unique and good position, where the location could be more controlled (since we have to use connections and services, at least occasionally, to actually use the device and some services). IP and network based location could be taken care of with VPNs. Metadata (language settings, used units etc.) and info given to services are given by the user. Which pretty much only leaves what the sensors tell - and the systems use to assign location - and that is the location data I was referring to.
Can those sensor outputs be made more user controlled, could location data actually be managed (as in: only give out my homecountry, show me at the gym when I’m out, jump my location around randomly in a 75km radius, increase error margin by 30%, add ghost signals from random hotspots, always show me moving etc.)? And there is a difference in not giving any location info (which seems impossible and sometimes blocks services) and giving false or less accurate info (which is for protection but also possibly enough for the service to work).