As a user, my impressions is that the latency involved with routing mobile data around is a lot higher (more like a second, even with LTE) than the latency involved with the physical transmission. Mobile data connections are generally more reliable than WiFi, but usually there is a significant lag before results are returned. Is that not correct?

This matches my personal experience with tethering being my main internet connection outside of work/office internet

Any omnidirectional microwave/RF network technology will have a short enough range that the signal propagation time is negligible for the purposes of human-perceived latency. This is a consequence of the need for cells to be small enough that they are only serving a manageable number of users. (Directional links like microwave backbones and backhaul can have latency worth caring about.)

Any latency you perceive comes from your packets sitting in a buffer waiting for their turn to be transmitted or waiting to be re-transmitted after a collision. The problem of queue management for wired networks has only had a satisfactory solution for a few years. When you throw in things like collisions due to the shared medium and interference from other users of the spectrum (less of a problem for cellular than WiFi), transmission rate and power selection covering multiple orders of magnitude, interference that can be problematic for some users but undetectable for users elsewhere in the cell, and the almost complete unwillingness of the hardware vendors to sell anything hackable enough to do research with, it's clear that wireless will continue to suck for the foreseeable future. However, a lot of these problems get a lot easier when you're working at a frequency that can never reach from the inside of your house to the inside of the neighbor's house.