You're making the mistake of thinking of Hubble's law as if it were describing the motion of a single point across a range of distances, whereas it describes the state of motion of different points at a moment in time.
Start with the moment when an initially expanding universe is decelerated so that it's neither expanding nor contracting (it's at the inflection point, just before the contraction begins).
At this moment, every recession velocity, of any distant point, is zero, and the Hubble parameter is zero.
Then the points start approaching. The point A at distance d from the observer is accelerated inward by all the mass that is contained within a sphere of radius d. Point B at distance 2d is accelerated by all the mass within a sphere of radius 2d. I.e. points further away are accelerated more, so the velocity they gain is greater. Just as Hubble's law describes, only with the Hubble parameter now going negative (which only means the direction being reversed from expansion to contraction).
After some time, those same points will have moved closer to the observer, and will have gained speed. But they still obey Hubble's law, with points twice more distant having twice more speed. It's the changing Hubble parameter (in this case, becoming more negative), that reflects the increasing approach velocity of points A and B as they get closer together.
edit: I should clarify - it's how H is changing, since it'd be changing even in a coasting universe, where all velocities remain constant for all time.
