To avoid quid quo pro, parsed, sentence fragments I'll skip over commenting on all your point. I'll conveniently ignore the put downs and decline the forum wager (really?, a forum wager).
I'll agree that we're both idiots.
The Google Adwords Keyword Tool is not an accurate reflection of what most people want in a domain name (as measured by the people buying domain names)
This is where we agree. The Adwords Keyword Tool has not been developed to allow people to make judgments on buying a name. It is a tool for customers to supposedly make more educated and informed decisions on advertising and marketing campaigns. It has sadly been crudely reverse engineered to provide domain value feedback and Google's tool has been picked up by everyone as some sort of evaluation Oracle. In a pure search-term = keyword = predictable traffic interpretation it probably, as you said, has more value. I don't believe that the numbers should be the basis for making a decision. Where we disagree on this tool is that I believe there's no harm - and there is in fact value - in using it as part of an evaluation.
The advice to ignore it totally is not sound for anyone who is looking for something analytical and something statistical to base their domain decisions on. When you say skip google analytics you are potentially ignoring a very important aspect of what potentially makes a domain desirable. That said YOUR personal experience may well make it redundant - no doubt you could have an idea whether a term is highly sought after and have a feel for CPC without even looking. With experience the reliance would be less and less. For non English speakers in particular you can at least avoid buying nonsense which would save countless 1,000 of bad registrations.
When comparing 700 searches to 3500 searches does it even matter?
Outside of some ridiculous long tail search terms that people should not hang their hat on... this seems to be the sweet spot that most domainers are working in when they are not buying expensive names in the aftermarket (a speculative guess based on what I've seen and not statistically measured). For localized terms MiamiLawyer.com vs. generics Lawyer.com (granted an extreme example) it is important because your searches are inherently more market targeted. Therefore the tool doesn't provide answers but it can give you some insight. That said - purely looking just at EXACT searches and correlating to value? In my opinion this doesn't work.
I don't think this view is too different from yours.
Where we differ is on the analytical value and proof provided by robust reporting. I don't believe there is robust reporting, and what there is certainly provides nothing for the idiots (include me if you want) to get an advantage from in an analytical sense and for certain segments of the market you're still looking at a keywords being important. While this is the case you'd be a fool to not make it part of your consideration.
Many sales ARE keyword driven contrary to your point: Headphoneamps.com, propertymanagementcompany.com, over50lifeinsurance.co.uk all scroll on this screen as recent sales over X,XXX.
Many are simply always with value LLL.com, etc.
But what the prices don't really reveal is: Why are the names valued to the buyer what they are? And the just as important, why did the seller not sell for less.
Do some have traffic? Do some have meaning? Do some belong to a marketing campaign? Are some more defensive? Are some true business names? Some sellers ALWAYS sell high even crap names - and this people still struggle with. They don't get certain market dynamics based on the sellers position, liquidity, reputation and sales platform. Most importantly, and what I didn't say but thought would be understood, is that your "robust reporting" is missing a HUGE volume of sales data. I would guess that under 10% of sales are reported. Hardly robust.
That said I think where the divide is that we are trying to fit one solution into two different markets:
Consider NFC names. There are a number of NFC keyword sales but looking through history of sales you can see a dearth of RFID and other similar technology sales despite large search volume. I would barely recommend buying many .COM with high search let alone a .NET. If you shift to the more brandable Tap/Wave/MobileWallet names then you will see that there are nice sales despite the search numbers.
That analysis would 100% support your view that the comparables is far more useful than any keyword search numbers.
So I don't buy into discarding one vs the other but understanding the value that each source of data provides. I think MWZD has it right though - if you DO use the keyword tool you have to account for whether you are .COM, .NET, .INFO. You would need far higher numbers to justify a .NET gamble based on those numbers alone. I would not even bother with a .TV, .INFO based on those numbers alone. I also wouldn't use sales figures to determine what is and isn't valuable but to simply provide input.
Finally (if you made it this far which I wouldn't blame you if you stopped because I'm rambling):
A forum wager? Really?