For purposes of going deeper into this, I think it's the question of determining the generic dramatic greatness of a film from the attributes that contribute to a profound sense of catharsis in the synthesis of action, character and theme. Noir skews the emotional landscape of film as it attempts to be a more "brute actualization" of feelings/actions, and there might well be two litmus tests for "greatness," the first being how well the brutal arc escapes from the accepted ("acceptable"?) dramatic precepts, and the second being how well these two arcs are blended and integrated into the flow of the film's action.
As for the scale of an effort to more systematically analyze the films from (or closely attached to) the Noir Top 25 list, I think it might be more useful to proceed initially with smaller steps but on a larger scale. A close reading of the ten films in the Top 25 "switcheroo" AND a look at the seven films you've selected would produce an analysis demonstrating the range of similar dramatic strategies and the effects imposed on the films in order to structure the overall dramatic flow, which will show why two of those seven films are significantly more successful than the other five.
That effort could also produce some more robust guidelines for a more large-scale method for systematically applying quality precepts to films in general and noirs in particular.
But an effort that I think ought to be done first, one that could then be used iteratively as the more detailed approach is applied, would be to take those 102 films from the Top 25 list that you ID'd as "excellent" and see if a more cursory approach to separating them into three hierarchies or tiers of excellence can be done. It would of course be best to do so by setting aside any access to the actual results of the poll other than its having created the list of 102.
See if you can perform such an exercise, and report back as to what those lists look like, and what criteria you were able to apply, building from the ideas for such that you've already referenced. I think that could prove very interesting, both for the lists themselves and what the exercise might contribute for further, more detailed forays into this topic.