Here's what a quick google search turned up.
This link [0] shows that US police kill over three times as many people, per capita, than Canadian police. However, I think there are other issues, like easy access to guns, that affect this stat so this isn't really a very useful stat for our discussion.
This group [1] has been collecting info on arrest related deaths but a quick look only showed me links to how they collect data and not the actual data. It also looks like they stopped their surveys in 2012.
This paper [2] has some of the most relevant stats I've found for our discussion and covers use of lethal force by police between 2009-2012. From their numbers, 83% of victims were armed or believed to be armed (this includes toy guns, etc) and over 93% of the deaths were from firearms. If we assume that the 7% of deaths were the police didn't shoot all apply to unarmed victims then 10% of the people killed by police were unarmed and shot. Their stats also show LEOs being killed in less than 2% of these incidents. Their numbers do show 9% of these cases resulting in an injury to LEOs but we don't have numbers on non-fatal police caused injuries so I don't want to compare apples to oranges.
[0] https://www.prisonpolicy.org/blog/2020/06/05/policekillings/ [1] https://www.bjs.gov/index.cfm?ty=tp&tid=82 [2] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6080222/
Unarmed in the context of this paper would only be as it applies to deadly weapons. The suspect could still pose a threat with an alternative weapon (such as a pipe) or even pose a threat with bare hands too. The percent of people killed without posing an immediate threat would be more appropriate, and was 5% (would like to see the details of those with prosecution metrics). I would guess that 5% that mostly fits in the 7% you mentioned (which includes cars, suffocation, falls, etc), but unfortunately the paper doesn't seem to get granular enough for that. So I don't think we can draw a good conclusion from this paper that there were more civilians unjustly shot by police than the police being shot by civilians (the paper leaves out non-fatal incidents too). On that note too, the 2% officer fatality rate you show would not be fully encompassing. Meaning, it would be missing encounters in which the suspect killed the officer without also being killed.
Every building search is considered high risk. It's not an extreme risk, which is why they didn't execute it as a no-knock with SWAT, but there is still risk. The actions taken appear to consistent with that level of risk.