You mentioned you already tried a different density mesh cut-off, that is a good idea, but maybe an even higher value is needed, 200 is not extremely high and when applying pressure, you may need a very detailed description.
Maybe I was not so clear about the point of rerunning from the 50 GPa geometry. It is possible (perhaps not very likely, but possible) that starting from the zero-pressure structure, your 50 GPa and 55 GPa calculations end up in different (local) minima. You could try to evaluate the space group symmetry of each result and compare, and as mentioned instead of starting always from the same, use the 50 GPa converged geometry when applying 55 GPa pressure.