Your mixes with Sparrow rely on the other participants true spend not being known. If LE are able to recover xpubs and know a large % of true spends, your anon set is likely much smaller than you think.