When identifying a line in a high-redshift galaxy spectrum, why is matching several lines more reliable than relying on one feature?