An interesting point, in my view, is that a certain "percentage" is permitted to "fail" in a statistical sense. In India (and this is outlined elsewhere in many of the state DOT manuals on compaction) we used an average of 10 to 12 "tests" on a single layer (using the nuke). The average had to be greater than "Spec + 1.65xSTDEV)" If this happened, the layer was passed even though single values could be lower. (I don't have the exact formulation in front of me at present - it could differ from that above) but it is analogous to concrete testing where the average of the last three test results have to be greater than the spec value but a single value could be lower by up to 500 psi low.
Too, the nuke should have been calibrated against, say, the sand cone (which most consider to be "more true") and the moisture content should have been calibrated to known values (taken from field tests). These aspects have been discussed in other threads on the issue of using nukes. Do you have details on the last time the nuke was "calibrated", it's repair history, leak tests, etc. These are issues that you can bring out - not that they are, in themselves, a touchdown, but bringing out these kind of things that should be in the testing company's SOP on QC testing might be appropriate to show doubt. Did the QC testing company have a SOP or QC Manual for such testing? Did (do) they follow it?
Was there questionable differences in the delivered material's grain size that might suggest different MDDs be required/used? Did you, as a contractor, carry out any QC testing for your own records? Most contractors on projects I have been working on for the last many years are required to carry out QC testing - the client or engineer provide verification testing for QA purposes. Was this the case? There are differences permitted from your QC and their QA verification testing that wouldn't "fail" your work. Did you carry out trial compactions, i.e., 4 passes = X% MDD, 6 passes = Y%, 8 passes = Z%?
All things to consider.