How the GAO ruled on the Northrop Grumman JLTV protest

The Government Accountability Office released its decision on the Northrop Grumman JLTV award protest this week.

Full version (.pdf).

The document notably contains effusive praise of the Lockheed Martin effort.

So why did Northrop Grumman fail?

Key graph:

(N)ot only was NG’s proposal rated overall only adequate with moderate risk under the ISM subfactor, but each of its JLTV-A, JLTV-B and JLTV-C configurations was rated high risk for reliability and maintainability (as well as moderate risk for design).

What accounted for the high risk for reliability and maintainability? I hope they're not talking about the diesel-electric drive system, because the Pentagon has since specified that it expects its eventual choice to use that system.

For each configuration element, offerors were required to complete a self-assessment of the level of maturity for each of the above four considerations, assigning one of six levels of maturity for each consideration as listed in solicitation attachment 24 (level 1 being the lowest and level 6 the highest). For example, attachment 24 generally defined the maturity levels for system design maturity as follows: level 1, completed concept system analysis; level 2, completed preliminary design analysis; level 3, completed system physical mock-up; level 4, completed detailed design analysis; level 5, fabricated system demonstrator; and level 6, tested system demonstrator. As another example, the maturity levels for system reliability maturity were generally defined as follows: level 1, reliability management program and failure allocation prediction; level 2, reliability prediction; level 3, reliability design analysis, in which the reliability design prediction is supported by identification of the critical components and physics of failure analysis, fault tree analysis, finite element analysis (FEA), and dynamic and/or static design modeling and simulation have been performed on the identified critical components; level 4, component level testing; level 5, systems integration testing of a system demonstrator; and level 6, system level testing of a system demonstrator to validate the system level reliability. As with design and reliability maturity, attachment 24 indicated that, for both system maintainability and C4I maturity, fabrication of a demonstrator was necessary to achieve level 5 and testing of the demonstrator was necessary to achieve level 6.
(bold emphasis added)

I'm admittedly no expert, but this response from the GAO makes it look like the JLTV evaluation process was poorly designed and executed. Suppose that diesel drive is reliable and you can work up a prototype fairly easily. You meet the basic requirements and you get the reliability of a traditional diesel system. Come up with something new and you take the hit for the difficulty of crafting the prototype as well as for the reliability problems of the new system.

But if the system is eventually expected to be of the latter type (such as a diesel-electric drive) then the reliability displayed by the diesel system is effectively irrelevant.

I sure hope I'm missing something, here.