rpc generatetoaddress doesn't always clear mempool

This is an intermittent issue and I just wanted to start tracking it. I'm doing auction tests with Bob as a full node, and connecting an SPV node as a secondary auction participant. Every so often when I `sendopen` or `sendbid` from SPV to full node, then `generatetoaddress` on the full node, the new block DOES NOT include the covenant transactions. I can confirm they are still in the mempool with `hsd-cli mempool` and usually after a SECOND `generatetoaddress` command, they will get picked up by a block and I can proceed with the auction.