Disk usage discrepancy?

You paid for what’s confirmed by your node and customers - it’s an average disk usage graph on the left. The right part is shown from the databases too, but databases are updated only when the filewalker is completed its scan for each satellite.
I also hope that you removed decommissioned satellites too (How To Forget Untrusted Satellites).
The best way can be to reduce the allocation and disable a lazy filewalker, to allow it to finish more quickly. However to completely clean your node from deleted data it should work like this several weeks.