It means that you commit so frequently (or generate redo so slowly) that the background write threshold (1/3 of log_buffer) is seldom the trigger for a LGWR write. Also log file sync waits are a little faster than log file parallel write waits, which means that for some syncs LGWR had already begun the write when the process began to wait for it. Given the fact that background writes are relatively rare, that suggests that there are DBWn syncs. You can check that with the APT script dbwn_sync_waits.sql. DBWn syncs could be due to overzealous checkpointing, or too small a buffer cache.

In my db, the sync cost ratio is well under 1.5, but average log write size is very different of background write threshold. What it means?

   Average Log     Background
    Write Size Write Theshold
-------------- --------------
          5632          87552

Sync Cost Ratio
---------------
           0.95