That is a great question - this is an account of what happened but not why.
Suspect the theory of alignment with the river from flinders street is what set the departure from the generally north-south/east-west grid. Assuming that then the change to a generally north-south grid was apparently just planning orthodoxy - hence the difference.
If you look at other early areas, Sandridge (Port Melb) was aligned with the coast whole Sth Melb kept with the Hoddle Grid in alignment as well as orientation. Williamstown bent with the coast with no consistent grid orientation.
The greater Melb grid coincidentally (?) aligns with the angle at which the bedding planes of the Melbourne mudstone intersect the surface - I wonder if that made early tracks any easier to construct on that orientation - but probably just coincidence. Hmmm.