It's the high reliance on stealth that worries me too. It is possible to achieve amazing RCS reduction using CNT-based material. However, we have to remember that RAM design is a trade-off between (frequency) bandwidth and RCS reduction (or, more correctly, reflection loss) even using multiple layer CNT material. Here's a set of graphs that illustrate my point. I should point out that the test material involved here IS a CNT layer, but is NOT that used in JSF. It's just illustrative of the properties.
The point here is that we can (in this case) get a -25db change in reflectivity (which is a huge reduction), but only in relatively narrow frequency bands. Clearly, additional RCS reductions can be achieved by form design, ect, but even this tends to be quite directional (it varies with aspect angle). The upshot is that you design to minimize reflectivity in certain directions and at certain radar frequencies. You cannot put on a cloaking device that is effective at all frequencies, at all aspect angles.
So, we seem to have here an aircraft that should be able to use its stealth features at range to reduce significantly the probable detection range of an enemy fighter equipped with a conventional, current generation radar. The obvious tactical advantage here is to obtain first A-A launch before the enemy have a probability of detection. So far, so good.
By definition these will be relatively long range shots (maximizing first shot capability) and, therefore, not maximum Pk (especially as JSF isn't the best platform to accelerate and climb to maximize energy at launch) and may be expected to achieve a kill rate of, let's say 25% - highly dependent upon dozens of factors, but let's give it the benefit of the doubt and assume 25%the remaining shots defeated kinematically or by countermeasures. We've achieved 2 things, reduced enemy numbers and disrupted them, probably causing them to manoeuvre. (NOTE: it is possible to have a much higher first launch success rate, but it would require the enemy to be unaware that they had been targeted).
But what next? The JSFs either keep closing on the enemy to prosecute the attack or attempt to reset. In the former case, even after an f-pol manoeuvre, they are increasing their distance from the carrier (combat fuel) and decreasing range to the enemy (and, therefore, increasing the probability of detection). So, we are rapidly moving to a situation where JSF's LO advantage is being lost and it's lack of manoeuvre disadvantage is becoming much more crucial. As we reach the stage when they're exchanging high Pk shots, the airframe/engine/fuel combination could place JSF at a significant disadvantage.
The reset option can be dangerous as the enemy could still be closing in range and at some point the JSFs have to face up to them, costing energy and time and exposing themselves at aspect angles where the RCS may be considerably larger.
All highly hypothetical and this is just one scenario. I deliberately haven't tried to predict an outcome or compare performances with specific threat types. I could offer others that would look much worse, but I've gone on long enough.
My conclusion: we're putting an awful lot of faith in LO on a platform that can be outclassed in CERTAIN other areas even by SOME current generation fighters. LO may give them a great advantage for penetrating on day one, but will it be enough to ensure operational effectiveness thereafter?