@Mishari: I remember we had a video chat with Dietmar a while ago. Can his house-number validation tool used for this purpose as well. That way it would only list not processed data, filtering out existing/correct one.
Splitting the data into provinces could make it easier for some as well.
If the file would be an online document (eg google docs) mapper can mark ranges as processed. Still an automatic filtering would be more reliable and less manual work.
I agree with you that we should refrain from automatic import and use it as a chance to also double check the spelling/location of nearby temples and villages. Even if the process takes a bit longer that way the quality we have at the end will be much higher.