Currently when mongoS chooses an atClusterTime for queries it knows targets only one shard, it will choose the latest lastCommitted opTime timestamp it has seen from the shard. This will eventually be changed to use the latest lastApplied opTime timestamp to leverage speculative snapshot behavior, but until then, to enable testing of transactions on single shards with enableMajorityReadConcern=false, the computed time should instead always be the latest time in the router's logical clock. This matches the existing behavior for writes and for reads that target multiple shards.