Did you clear the wait statistics after you lowered maximum degree of parallelism to 1?
But there is really no valid simple, universal rule of thumb about how this setting should be configured, other than to measure the impact of any changes you make. In addition to the above URLs, I'd recommend reading this post by Paul Randal.
I'd also suggest reconsidering focusing more on the other top wait types -- I often find that CXPACKET waits are a red herring and that they're occurring because of complex queries that get parallelized into simple pieces of work that complete quickly and larger, I/O-bound pieces of work that take a long time to finish. The CXPACKET waits show up when the small, fast pieces of work complete and wait for the larger pieces of work to complete.
If you're looking at tuning specific queries, the execution plan should show you what parts of the query are taking a long time to finish, and then you can tune to address the problem. Out-of-date statistics or poorly chosen indexes are often the culprit there.