Figure 2: yes. It is the cleanest single datapoint in the dataset and the like-for-like basis (iOS-only, US-only, seven full days) survives the obvious objections. Keep Threads out of the bars; the footnote is the right place for it.
Figure 3: yes, and it is the most important of the three. The right panel answers the cannibalisation question directly and the index framing makes the 59% decline and the >3x combined expansion readable on one axis. Agree that App Store rankings belong in the annotation, not a standalone chart.
Figure 4: yes, with two caveats. The left panel's DAU/installs ratio is a launch-week diagnostic, not retention (denominator is still growing daily), so label it that way. The right panel mixes two units; keep it as two adjacent bar sets rather than a dual-axis chart. Four of the six "metrics to watch" (tasks per user, repeat transactions, combined MAU, international conversion) are not observable in Sensor Tower yet, so the box should be labelled as company-disclosure asks.