A is a neutral expression.
B is a somber event where restraint is required: A is appropriate.
B is a happy event or perhaps a close family member: A then B implies he's cold, underwhelmed or disapproving.
B is some random, inanimate object: B holds no new social meaning of A, unless B is symbolic or a plot Macguffin.
It's the same A, but it's the clip B following it that rewrites socioemotional history of the viewer.
At a higher level of abstraction, it and other socioemotional back-rationalizations imply most people allow the future to rewrite their impression of the past, making the past not entirely cast in stone. People can easily turn on each other and then forget about it simply because they choose to rewrite history in their minds, twice.
(FWIW, I have ASD as well, and for me the expressions on his face are meaningless.)