We are seeing a very similar issue. In our case, AdServices/self-attribution data is significantly higher than Apple Ads Reporting, but the discrepancy is almost entirely on Click attribution. Impression attribution is much closer to Apple Ads view-through installs. This pattern is important because our internal ingestion logic does not treat Click differently from Impression, except for the claimType value returned by Apple. If this were just a duplicate upload or backend counting bug on our side, we would expect Click and Impression to be inflated in a similar way. Our current hypothesis, which we believe is highly likely based on your experiment, is: attribution=true from AdServices is not a one-time consumed event per install/reinstall cycle. It appears to behave more like an attribution-window-level match. For Click attribution, as long as the device is still within the 30-day click attribution window, subsequent installs/reinstalls or app opens may still return attribution=true. For Impression attribution, the window is only 24 hours, so the same repeated-attribution effect is much smaller. Apple Ads Reporting appears to apply an internal conversion/installation attribution logic that is not exposed in the AdServices first-open payload.