you need to the following
Best method is to make sure special characters are controlled at the source system level. But in real time, end users do make mistakes hence better include them in RSKC. But RSKC can hold only 73 characters. So, make your system is "UNICODE" enabled. Unicode must be made at the source system level as well as BW level too. This will help us a lot.
Correcting huge volume of the data at the PSA level is time taking. when users are waiting for the report you cannot stop the load for correcting the records in PSA level.
looking at the PSA you can filter the records in DTP level. But if there are multiple records and multiple special characters you will not be able to do it manually.
In our case, we actually went for the custom FM to replace the special characters with space. But it did not help in longer run. So we made our system UNICODE hence we now have support of Arabic, English and German + we maintained some characters in RSKC