Psychologically, I tend to agree with your assessment, but I have a beef with their math too.  I believe value (if calculated correctly using a pythagorean model) accelerates with increasing divergence from average...a mega-star isn't worth the linear gap in run value...it's worth the inverted pythagorean gap...which has a quasi-exponential lapse rate.  A true ace pitcher isn't worth 5 wins to the big #2's 3 wins.  He's worth 10 wins to the number two's 3 wins.  The same happens for the hitters (but not quite as severely because there are nine batting order slots so the payday is split more ways).

PCA credits Pedro Martinez (2000) with being worth roughly 14 wins above the zero-value margin (about 11 wins above average), whereas it credits Josh Beckett (2007) with being worth 6 wins...a still-outstanding mark, but not the kind of pitcher you attach multiplicative value to.  The ace-caliber seasons POP out of the PCA ratings...a lot more than they seem to do with WAR.

...you can read the PCA manifest...I don't know if you have time to go through the whole thing or not...it's 75 pages (James' explanation of Win Shares is similarly lengthy...hard not to cover a lot of pages with verbage when explaining something as complicated as a full uberstat metric system)...I produced the explanation for anyone who wanted a copy.
To clarify the main idea of the system though, I use Pythagorean theory (hence...Pythagorena Comparative Analysis) to set the margin for performance, rate teams and players relative to the margin, and give significant additional value to performances that imply a forced sub-marginal performance from the opposition.  The system still uses the linear-above-margin approach for dividing credit among players (just as James does), but my marginal baseline is significantly higher than his was and the concept of supermarginal performance (adding a second value enhancement to player performances that imply the opposition is submarginal when facing a given player) places additional emphasis on superstar caliber play.  Think of the supermargin as adding the negative wins a supermarginal performance creates in the OTHER team to the team being analyzed.  I call this quasi-exponential because the supermarginal value is pythagorean, so the further above the supermargin you are, the higher you rise above your peers.
I'm not sure I discussed the supermargin in detail in the manifest...I was mostly laying out the logical foundation for my approach to sabermetrics...but I still think it's worth reading if you are actually interested in how I did my work.  PCA is a look into a little time machine for me...some of the methods I used were a little primative compared to the ideas I have today (I last published PCA after the 2004 season...it's been five years).  But when I read back over it, I am still proud of all of the careful thought I put into it.

I have short-hand calc methods like James does that I can whip out to get rough estimates.  I think Bedard was above the supermargin in 2007 but not since he's been a Mariner (his walk rates are too high in 2009, although that's probably caused by the injury implosion in June/July).
The 2003 Mariner outfield defensive unit was supermarginal (I don't rate the supermargin for an individual fielder since no one fielder can create sub-marginal performance form the opposition, but a whole defensive unit (Pitchers, catchers, outfielders, infielders) can, in rare cases, do it).  The 2009 Mariner outfield was roughly as good as the 2003 outfield, so it's fair to guess that the '09 outfield defense was slightly supermarginal (which is why the whole team has such a stunning DER despite Betancourt/Cedeno playing short most of the year and Lopez being only average-ish at second).  None of the Mariner bats were supermarginal this year, I don't think.  Ichiro, as good as he is, is typically scraping along the top of the standard performance range...you really have to whip out a 170 OPS+ or something silly like that to break into the game-changing dynamics.
Aardsma probably was not supermarginal either with his walk rate, and he was the best bullpen representative.
I plan to completely recalculate PCA with some enhancements this off-season.  It won't be the full fledged idea I've been working toward - time doesn't permit me to do all of what I want to...but it will address some of the flaws with the old system and provide concrete ratings for the last five years.

You're right that the Rangers had huge holes at a lot of other positions and it killed their overall wins...but an uberstat like PCA should SEE that (and it does...the Rangers' pitching staff as an entire UNIT was very close to the margin and their fielding units weren't much better).  You can track supermarginal performance easily enough.  The question is how do you decide what that supermarginal talent is worth.  My first guess was to place the supermargin at the exact opposite side of the value spectrum from the real margin and track supermarginal wins linearly up (on top of normal marginal wins) in the same way that I'd track sub-marginal wins linearly down from zero.  Maybe that's not right, but it's better to guess that than to ignore the supermargin altogether as all other metrics do.

I don't assign a super-marginal score to specific defenders individually.  Units get the supermarginal bonus and they split it between all of the players involved in the same way the unit's marginal value gets split...basically the super-marginal credit gets turned into wins at the unit level and the individual fielders are just splitting a larger piece of pie.
The team defense does help Felix, but DNRA > FIP in accuracy and value IMHO and Felix (using the pitching analysis in PCA, which is essentially a slightly simpler version of DNRA where certain assumptions about the rate of XBH, GDP and rare events that aren't in the pitching record other than PBP data) are applied) did a lot on his own to prevent run scoring that isn't captured in straight DIPS.  I do not know if it would be enough for him to be well supermarginal...there are usually about 1015% of all pitchers in any given year who are (and Felix would suely be among that group), but most of them are only getting slight boosts.  The slope is steep (which I consider a good thing).

Of course Felix's actual WAR in '09 was 6.9. His season was worth $31mil+ in last years market using FIP and he could get better. 5 WAR is an extremely conservative '10 estimate IMO.
At $18-20mil a year, Felix is actually quite a bargain. You'd need pretty wild return for that kind of guy.

...is in itself conservative if you assume some supermarginal performance and recognize that he was better than his FIP (accounting for the gains he created on batted ball results above the Mariners defense)