There is a very broad middle ground here, based on paradigm of localized use. It doesn't get rid of the need for large, non-solar power plants, but it does offer some relief.
What you've done is made a case against solar cells in their entirety. Your set of "narrow exceptions" misses a whole range of possibilities.
For example, it's possible to run a home air conditioning unit using power generated by roof-top cells. The highest demand and best opportunity to use the system occur at the same time: during periods of sunlight.
It's true that there are cost/energy return issues. But those are not unsolveable.
You just couldn't do it and make any kind of economic sense (without government subsidies, that is). Let's look at an example.
1. We have a 3,000 watt window air conditioner unit running 8 hours per day for 180 days per year at a utility rate of 15¢ per kW-hr. Cost per year = $648.
2. When calculating PV output we will use no loss for night time, sun angle or dirt. The starting value of 790 W/m^2 is reduced by PV loss (85%), packaging loss (20%), heat loss (15%), inverter loss (20%), leaving 64.5 W/m^2. For 3,000 watts, that's 46.5 m^2 of PV panel array needed. The area of the SP-150 (for instance) is 1.32m^2, meaning that you need at least 35 of them at $700/unit or $24,500. Plus another $3,000 for the 3kW inverter and another $5,000 for someone to install and wire this system. Now we're up to $32,500. I'll give you the building permits for free.
3. But for that same $32,500, I could have run the air conditioner off the power grid for 50 years! And this does not even begin to account for the time-value of the $32,500 I had to plunk down on day one to start this venture!
So maybe you could explain to me again just why it would be such a good idea to buy the solar cells to run the air conditioner?
--Boot Hill
The same roof collector surface that forces you to use air conditioning is what you propose to use to offset it?
An umbrella would be a cheaper alternative.