According to 6.2, "thus ignoring Extend is sufficient to disallow breaking within a grapheme cluster." However the sequence of Unicode scalar values (U+0600, U+0020) is considered a single grapheme cluster due to rule GB9, but the sequence is parsed into two words according to 4.1.1. While it would be ideal to not have sequences of Unicode scalar values that can be parsed into more words than grapheme clusters, I think it's more understandable if section 6.2 didn't explicitly state that this isn't possible.
