A better measure of market value would be what free agents actually get paid comparted to their average WAR for the previous 2-3 years.
I always applaud an analytic approach to decision making.
You seem to be aiming toward a forecast of WAR, at the time the contract is signed. I think the studies computing $6M or $8M (or whatever) per WAR are doing it in terms of value delivered, looking back after the contract is signed (so the jury is still out on recent guys). That is, they are asking the question, what do teams typically get for their FA money?
I think it stands to reason that front offices' forecasts are getting better and better, so as time goes along the actual value delivered (at least insofar as WAR represents "actual" value) becomes a better and better proxy for the forecasts.
But even in trying to come at the question from the forecast WAR for these players, a simple average of previous years seems awfully far from what teams probably use. I don't think you would use that kind of average for a product life-cycle study, for instance, unless at a very particular stage (probably mid-life) - you would draw dangerous conclusions if in the early years, or if at the end in maintenance mode when you're trying to wean customers off. Ballplayers, as a "product", surely aren't static enough in their "life cycles" to try that with.
With the horizon you specified, I would weight the most recent year more heavily than the prior year, and much moreso than two years back - almost ancient history in some cases. But I would also factor in growth or (more usually, for these free agents) decline relative to age. Injury risk also comes into play. It may be that teams are (in the privacy of their own processes) putting a very large downward factor on their forecasts over the life of the contract they intend to offer. That would move things in the opposite direction from your conclusion, since it makes the denominator smaller.
Another thing that makes the methodology difficult is that players' "accurate" forecast of value, by whatever procedure you think best, is probably in the middle range of what teams will compute for themselves; and it's highly likely that whoever computes the highest value will make the largest offer, and in turn the player is highly likely to accept an offer very near the top of the range. This of course would move things in the direction you suggest, as it makes the denominator larger.
All things considered, it's hard to approximate teams' forecasts with a simple average. Beyond hard, I think - misleading, or even not useful. We'd be dividing $$/WAR using something basically unknowable.
Cespedes, the first guy on your list, seems like a good example. He was a highly sought Cuban free agent in 2012, and when he became available again in the 2016-17 offseason his resume was a bit spotty. He was coming off a 2.9-WAR season (I'm using b-r.com) after a 6.2, decidedly his best, and he had missed a few games in August after putting in two full seasons the prior years, after starting off with two injury-impacted seasons. I could imagine some widely differing forecasts by competent professionals in the field. When the Mets prepared their eventual winning offer, it's not preposterous to think that another 6-WAR season could be expected. But, he was already turning 31, and for the four-year contract the team was contemplating, a decline could be expected. An injury during any of those four seasons could further harm the value delivered in that season just due to his absence, while also perhaps accelerating his declining ability for future seasons. All in all, just spitballing here, a four-year WAR of (6,5,4,3), times an 80% chance each season for not having a really serious injury, comes out to only about 14 WAR over the life of the contract, or about 3.6 a season, rather than the 4.4 you came up with. That comes out to around $7.5M per. If I did the math right. Again, I'm only spitballing.
Now, given that the Mets won this sweepstakes, it's fair to assume that most other teams* came up with a lower WAR estimate and made commensurately lower offers to the player. Unless you believe the Mets are super geniuses and have a unique ability to forecast future value, it's very likely that the actual value he returns will be more in line with the crowd and thus lower than they thought they were paying for. Ergo, the cost per WAR will likely be higher than their forecast (and perhaps this guesstimate). That's the so-called Winner's Curse in any free market, right? (Not that MLB markets are all that free. )
I'm not going to invest the time with my rinky-dink eyeballing methodology, on the other players you listed, but you probably see my point, that trying to infer MLB forecasts of WAR is harder than just averaging some recent seasons.
* Even if you rule out the small market teams, there are enough other deep pocketed teams to make this line of thought work.