Here's at what I was attempting at getting at with the disregard to certain qualities work.
These days there's a popular trend when simulating things to simulate every possible mechanism we can imagine. Those who think that way would agree with you. Why would you ever make a flat Earth model when everything is eventually going to make its first flight on a real rotating spherical-ish Earth?
This approach works great until you come across real development or computational limits. The cited paper is from 1988. Computers were much weaker back then. For perspective, the Cray Y-MP was sold that year. Its peak performance was 333 megaflops. She cost $15 million dollars. Contrast that to today. A Geforce GTX 1070 is capable of 6,500,000 megaflops (6.5 teraflops) and has a price tag of around $400.
In those days, you didn't waste computational power on frivolities. It turns out that for a vast array of aeronautical problems, the effects of a flat earth vs. round are minimal (much less the effects of rotating vs. not). If you're shooting a shell 15km, and need it to land with pinpoint precision, you need all that extra complexity. However, many aero problems include a guidance unit which would address any error due to Coriolis effects or the spherical ground the same way it would handle any other errors. It'd simply see it wasn't on the right path and make a correction. The other sources of error here, such as winds, play a far larger effect in deviations from a flight plan, so all the rotating and spherical effects can just get lost in the noise.
Even today, we still make flat Earth models. The reason is not computation time, like it was in 1988, but development time. The more things you model, the more things you need to develop, verify, and maintain. If a particular problem does not call for advanced models, why waste budget developing and maintaining them?
