Sigh.
1. Don't reference Other Countries nuclear programs. This is the United States,
My $4/W figure was the estimate for new United States reactors, according to the interdisciplinary MIT study The Future of Nuclear Power (the 2009 update).
Referring again to the MIT study, they explain in detail what goes into their cost models (the 2003 full report, appendix 5). It encompasses EVERYTHING - the entire plant (steam turbines and all), the operating costs over 40 years of operation, 40 years' worth of fuel, the decomissioning costs after those 40 years, the waste disposal cost under the current 0.1 c/kWh DoE fee, etc. The TOTAL cash flow is estimated at $4.5 billion (nominal) during the construction phase - see the supplemental paper Update on the Cost of Nuclear Power, table 6A (this doesn't include the financing costs - go down to 6C).
Of course, what's really interesting is the levelized lifetime cost, per kWh. The MIT study estimates this at 8.4 c/kWhe; I've surveyed a dozen other such levelized cost studies on my blog. Feel free to follow the links and read up on them.
By the way, the NRC fees a very tiny part of costs - currently $4.6 M/year, out of of the MIT estimate of $56 M/year of fixed O&M costs (for a 1 GW plant).
6. Definitely not an engineer. Megawatts are always comparable, they are absolute quantities.
A MW produced by a wind farm is the same MW produced by a nuke.
Nameplate capacities are incomparable. They represent peak power generation; but some power plants always operate at full power, and others operate intermittently, hence the energy yields (integral of power * dt) are completely different.
Yes, while wind provides a smaller percentage of it's capacity factor when compared to nuclear, that can be (supposedly) be defeated with large numbers of geographically dispersed wind farms.
No, that's a fallacy. 1 MWe of wind (nameplate capacity), at 30% capacity factor, averages 300 kWe (averaged over long time periods), with an instantaneous range of 0-1000 kWe. Adding together a thousand such (identical, independent) turbines gives you an average of 300 MWe, albeit with lower statistical variance - smaller fluctuations.
You are conflating two separate issues. One, is that the average output of a windfarm is a fraction of its nameplate capacity. Two, is that the output over time has very large variations. See? They are separate problems.