This is a reasonable request. But I need to think of a good way of doing this. Perhaps each translation defines the list of original messages that are known to NOT be translated. Then, those keys would be ignored when the tool runs and would not contribute to the percent complete score.
One suggestion: Could you modify the report so duplicate keys would be considered as good keys? Nowadays, spanish and finnish translations, for example, are complete. However due to some keys are similar in english, they are considered as duplicated and completion percent is lower.