So your stats may or may not show a bias. After all, same number of whites and blacks were stopped and killed. And since you are stopping a higher percentage of black people (10 percent as opposed to 2.5 percent), you are more likely to encounter the violent black folk is higher, so you'll get more dead cops as well. All in all, this sample bears out your numbers.
But that ignores selection bias. It's not actually true that you would sample randomly from both populations. Police would obviously try to stop people who seem more su ious. It's not easy, but it's theoretically possible to pick 50 truly violent white people in this scenario. It's not possible to find 50 truly violent black people, as the population just isn't big enough. So you're going to have at least half of the black people stopped being innocent while you'll have fewer innocent white people. So assuming that those five deaths are actually random (and boy, it's starting to look like it some days), you'll get a larger percentage of black people who are killed even though there was never any reason to stop them in the first place.