Algorithmic amplification is an entirely different kind of distribution dynamic that isn't covered by your analysis. Algorithmic amplification is inherently
non-neutral, so AA is already biasing distribution. Neutrality towards some content is to not algorithmically amplify it. At best you can say Facebook was neutral towards this story. But that doesn't have the same ring to it.
Do they have a duty to be neutral with the application of their algorithm? Obviously not: such an algorithm must discriminate on content in some manner. There is no such thing as "neutral amplification" in this context. Do they have a moral duty to not consider political content in the application of their algorithm? Obviously not, as a private company they are free to bias their algorithm in whatever manner they see fit.
Can we presume there is some kind of social contract that should constrain the manner in which they bias their amplification? Perhaps one could argue for this. But this social contract does not extend to foreign actors attempting to sway an election. If Facebook had reason to believe that this story was a part of an intelligence operation--and they did given the FBI warning--being neutral towards this story is not breaking any social contract. Facebook has no duty to allow their product to be the vehicle of a foreign psyops campaign (or a local one for that matter).