These don't seem to be controlling their numbers for actual proportion. They're comparing direct %s without considering the higher proportion of white Americans. If there were no bias then the %s shown should match the relative population sizes.
For the black case, 29% weren't attacking, 15% were unarmed. What's to say the 15% that were unarmed weren't attacking?
There's no way to draw any meaningful conclusion from this data, because it's crap data.
What would be useful? Relating scenarios where someone shouldn't have been shot, like the percentage of unarmed people that were shot and not attacking.