You bring up some good points.
Just about all the power curves aren't linear, so it's not a 1:1 relationship with speed to power. That's why it's better than just looking at speed in order to measure your improvements.
You're also right about spokes not having a linear relationship with power. I wonder how many watts that would be. Remember, the best power meters are +/- 1.5% accuracy. So if you're power meter reads 300 watts, you could be at 295.5 to 304.5. And that's only if they are calibrated correctly.
You're also right about the inertia it takes to speed up and slow down different size wheels. We don't calculate this or even try. How much are you losing? I don't think much. If you're doing 3x15 minutes your wheel is already spinning so you're not losing that extra power to speed up the wheel. With the nature of indoor riding where you are pretty stead state this works out well. If we were outside, with constant accelerations/deceleration, this wouldn't work well.
After you figure out if you want to use a PowerMeter or use VirtualPower, then I agree again. It's all about doing the intervals and doing the work in a structured program. If we get you to ride more often and are less bored, then I think that makes it worth it for a lot of people.
Oh, and when you're FTP gets up around 300, you should just get a power meter because you are obviously a bad ass and into the sport