A custodial wallet doesn't quite explain the evidence provided by nutildah. If it truly were a custodial wallet where you don't control the private keys (like those from centralized exchanges or gambling platforms), you wouldn't be able to initiate individual transactions from those addresses, as shown in the example provided by nutildah:


The transaction in question has three shared inputs and two outputs, one of which returns funds back to your wallet (change address). This suggests control over the private keys.

Now, if you're talking about a different kind of "custodial" wallet, like the one on blockchain.com where you do control the keys, then it wouldn't make sense to only have some keys and not all of them, right?