You could try seeding. From a new protein preparation of the same type, set up crystallization conditions nearly identical to this. Scratch these crystals by rubbing a pipette tip over them. Drag the pipette tip containing the seed crystals through the hanging droplet of protein and reservoir.

The nucleation barrier has already been crossed so the crystals should just grow in size from the broken fragments.

I would also set up an optimisation screen with varying ammonium sulphate concentration and PEG3350, this might change the crystallisation from clusters to more individual crystals.

also you can try aging, where the smal crystals dissolve and incorporate in the bigger ones. Your crystals look like to be precipitate on a very fast way. try do add the precipiting agent slower in the solution. and dont forget the suggestion from Adron to seed the solution