Consider the boost cost alone. That would cost more than the sale of power from the satellite for its entire lifetime. It is an idea like the space elevator: Sounds good in principle, but totally unworkable in practice. Just forget it.

Suppose that the satellite has a mass of 10 kg per square meter, which is rather light for an extended structure. It costs well over $10,000 per kg to put something into geosynchronous orbit.

This is based on launching as we do it today, with a typical flight being ground to LEO, boost into GTO, then circularize to GEO. A lot of the cost of doing that is due to the upper stage(s) necessary to do the GTO and GEO burns. Depending on where you're launching, this can be a huge problem. The worst case example is probably the Russian Proton booster. It's capable of putting 20 metric tons into LEO. However, if you're going to GEO, you need an 18 metric ton upper stage due to the high launch inclination. The end result is not only an expensive upper stage but a payload of 2 metric tons to GEO. (BTW: the same booster can send 5 metric tons to Mars). For launchers operating at lower latitudes, the upper stage doesn't have to be so large but it still is expensive.

Here are some numbers for comparison from the Encyclopedia Astronautica:

Atlas V 551 (Heavy)
Payload to 28 degree LEO: 20,050 KG
Payload to GTO: 8,200 KG
Payload to GEO: 3,750 KG

For a satellite with a lot of solar power, you could use ion or Hall effect thrusters to go from LEO to GEO. You probably could even use VASIMR. It would take a while to get to the final orbit, but you wouldn't have to use an expensive upper stage. The propulsion system could be part of the satellite and could be used for orbital maintenance.

Suppose that the PV cells are 50% efficient over their entire lifetime. That gives about 650 watts per square meter. Now the antenna array must be pointed at the Earth while the PV array points at the Sun, so that they can not be mounted on the same structure.

This isn't an issue. Take a look at the Milstar military comsat. The first one was launched about 10 years ago. It has a solar array span of around 100 feet. The arrays, like all 3-axis stabilized satellites, are on a rotating mechanism to keep them pointed at the sun. The axis is parallel to the Earth's axis, so all it takes is a simple clockwork type mechanism to point at the sun. Due to the geometry of the Earth's inclination and the sun, a satellite in a low inclination GEO will be in sunlight about 99% of the time, with eclipses lasting no more than 72 minutes a day on the vernal and autumnal equinoxes.

Suppose that the antenna array has the same area as the PV array and has a mass of 5 kg per square meter.

There is no valid reason to make this assumption. Solar array size is driven by the energy density of sunlight. About the only things you can do to reduce the size of the solar arrays are to use higher efficiency cells and/or concentrator cells. The same doesn't apply to transmitter modules. Consider my old stomping ground, the Cobra Dane phased array radar. It was disigned in the 1970s and operated in the 1.215-1.4 GHz frequency range. Each of the active elements on the array face radiated 1 kilowatt of RF energy and was only a few inches across. If you use higher frequencies, the elements can be a lot smaller. Further, with modern electronics, each element can be individually powered with tremendous redundancy. This isn't to say that it wouldn't be challenging - there would still be huge issues with cooling. However, it suggests the SPS could perhaps be modular.

Now consider conversion efficiencies and losses. For each watt produced by the PV array, only about 50% (if we are lucky) will go through the wires, through the conversion to microwaves, through the atmsophere to the receiver arrays, through the conversion back into electricity.

Even of the cost of the satellite were ZERO, which it will not be, it would STILL be absurdly expensive.

Granted.

Add the cost of the satellite hardware: (PV array, antenna array, conversion equipment, housing for workers, supplies), the cost of assembly in orbit, the cost of maintenance, the cost of the receiving array, the cost of conversion equipment on the receiving end. If it was expensive before, it is even more expensive now. Look how much the International Space Station (in low orbit, not synchronous) costs. Multiply it by 10,000 and you get the cost of a single orbiting PV station. Realistically (given the usual cost overruns), figure $1,000,000 per watt.

Unlike many things, the cost of electronics has dropped dramatically for a given level of capability. Your statement, "If it was expensive before, it is even more expensive now" doesn't track with what has happened to electronics prices over the past 40 years. As for the rest of the costs, someone needs to do a real analysis instead of just guessing at the costs. One thing that could go a long way to reducing the costs is to avoid having the government build it like they did the ISS. When you have money to burn, let the government do it. If you want results, go to the private sector.

I'm not saying that SPS could be economically viable with today's technology. All I'm trying to do is point out what appears to be inconsistencies with your analysis.