I feel there is some problem even when operations like CTAS are NOT running. Response time is comprised of @ 15% service time and @ 85% wait time. Almost every commit is waiting for "log file sync".

I sincerely suggest you to check everything @ the disks where your online redo logs are hosted. Do your ever see at least one of the redo group status as INACTIVE ? How big are the online redo logs ? How many archive logs are getting generated in a day and during peak times ?