You might check the bottom on surface defects of the injection mold. I once had a problem with a cold piece of material (from the open nozzle) stuck at one side bringing flow completely out of balance. (Due to flow it can melt away , leaving only small surface defects.) Best way to investigate is to make a birth-shot: start with 10% shot-volume and gradually increase volume to see when the unbalance starts.
Other possibility:
Use sufficient pressure to maintain speed: If speed drops below a certain value, you will always have unbalanced flow: Slow flow generates less heat, increasing viscosity; fast flow generates more heat, reducing viscosity, e.g. unstable system. What is your injection time and injection mold wall thickness?

I have tried all that, and what happens is an even flow in both cavities but both short, so short it's usually half the product. If its high speed low pressure I get a similar result but even shorter, but both even. As soon as I try to increase slowly, the flow goes to the outside of the cavity only on both sides.