Go read the source and review the model page on huggingface and u will find thats not at all true for my implementation. The model im using provides a summary 60% identical to that of a human summary and >95% accurate in maintaining the original meaning.

Plus 99% is one hell of a an exaggeration.