Let's suppose you your buying a large scale solar system with actual money right now.
Utility system prices are $2.04/W in Q3 2013. http://www.seia.org/research-resources/solar-market-insight-... it's less now but that's close enough.
Standard warranties offered in the market offer 90% rated power output for 10 years and an 80% rated power output for 25 years http://www.solarenergyhome.co.uk/performance.php
With real world numbers averaging ~90+% of stated efficiency over 25 years.
1W solar produces ~8-10 wh/day * 365 days * 25 years /1000 kw per w = 65.7+ kwh over 25 years. 2.04 / 65.7 = 3.1 cents per kwh worst case. Panels don't instantly die at 25 years. So, actual lifetime output is guaranteed to be higher so costs are lower. But there are vary low maintenance costs and occasional physical damage so that's a reasonable estimate on average.
The time value of money is only meaningful if you want to compare your returns with other options over the long term there is no point in including inflation etc. As it's not like 1kwh will magically turn into more than that just by time passing.
I don't disagree with your approach in a general sense, but what I think you should understand is that this approach is not a true measure of the input costs along early points of a technology's adoption curve.
For example, it looks like I can pick up a high end i7 for $300 at standard consumer prices. It has the cutting edge lithography standards, the kinks in that manufacturing process are still getting worked out (wastage), they are still paying off the initial R&D investment, they are actively marketing & selling it, and it's paying for future R&D investment. I can get a nearly equivalent (roughly speaking) Celeron for $40 from the same site, and I didn't even price out ARM/generic fab cpus. The Celeron price (or more accurately, the average price of that Celeron over time) is much more indicative of the true energy input costs of the underlying technology at scale.
A manufactured product has a virtuous cycle (from a consumer perspective, at least) of high upfront cost for development dropping as less marginal R&D is required, at the same time as the production process gets further optimized/automated and economies of scale kick in.
Solar is still definitely in the early stages of adoption based on growth curves, and prices reflect a huge amount of non-energy input costs that are not truly representative of the EROEI that will be present when it represents 10, 20, or +50% of total energy production.
There are certain physical properties of PV cells that prevent efficiencies of +35% or so in standard unconcentrated single-junction pv cells. Because silicon is so ubiquitous, I suspect that we're already hitting points of diminishing returns in research/cell efficiency and production/process optimizations are going to start dominating until product optimizations are necessary to increase density. Right now, the biggest cost I'm aware of is silicon wasteage, which is the driver behind thin-film tech or alternative geometries (eg Solyndra) - I have heard of research into vapor deposition approaches that could greatly limit this. Hypothetically, if that works at scale, I can't really see a mass-produced 250w panel costing more than $50 (could be even lower assuming a market/industry for refined Si recycling from old/damaged panels).
Same thing would eventually go for concentrated multijunction modules that can hit +80% efficiencies, eventually, assuming that the non-Si junctions don't have material constraints (like HRE elements).
Now, I've gone on too long already, but let me say I'm not de facto anti-nuclear. I think light water reactor design is terrifying and I think existing waste storage policy is basically sticking fingers in ears and yelling 'LALALALA'. In order not to end the world with a climate catastrophe, I think it will be a bridge fuel and hopefully novel reactor design tech (eg CANDU) can help there.
However, the future is solar and wind power supplanting coal and nuclear, bio-petroleum from algal/renewable-electric input supplanting necessary high density applications (oil), and biomass methane supplanting NG for intermediate applications like peaking plant generation.
edit: Evidence of this happening already: http://www.seia.org/policy/environment/pv-recycling ... you've accurately noted that there are substantial subsidies, but they exist to accelerate the maturation of this source of energy. In that sense, though, they've been effective and relatively inexpensive.
Actually I claim there is virtually no difference.
Every single time I read an EREOI calculation I kept wondering "But he ignored that energy, and that, and that". After thinking about it I realized that money makes an excellent proxy for EROEI.
Every step in the process that costs energy also increase the cost. So the money essentially acts as "book keeping" for energy.
Everyone always gets mad when they encounter this for the first time, but think about for a while. Think about the basic cost for every single thing we make, and you'll realize at its core its always energy.
There is one exception that I thought of and that is pollution. Reducing pollution is a valuable thing even if it costs extra energy to do so.
Because at the end of the day we don't care how much energy we use, we care that we don't dirty our environment, and that we don't run out.
Your edit #2 is well taken. I will need to think about that.
My initial impression is that nuclear costs more energy (all those externalities you mention) than hydrocarbon fuel, but makes less pollution. So maybe it's worth it anyway.