I do analyses like these for companies sometimes. They almost always get it wrong. I'm surprised Google's report doesn't seem to do it correctly either.
The issue is rarely that people of a specific race are leaving faster than other races. The issue is almost always that turnover is high in lower level roles and that's where the bulk of your black / latinx population sits. You CANNOT look at overall attrition rates for a company. You must segment it by level unless your levels have equal mixes of races, and they never do. I'd bet heavily that low level attrition is just higher than usual. You can see it in their charts that attrition has increased across the board. It just fails to notice that attrition is going to be among low level employees and low level employees are more likely to be black.
In order to not have black employees leaving at higher rates than white employees overall, while black employees occupy lower level roles, you would need black employees to leave at significantly lower rates than their white colleagues, normalized to level.
edit: and I'll also note, for a growing company the more important number is hiring. They're at 8.8% black/latinx each up from ~6% each last year. That's very good. The overall representation has improved by ~0.6% since 2020. That's also very good. Headline is misleading. Google's making strong progress on its diversity goals.