You will need to use a SplitInBatches node with batch size of 1 before the MSSQL node. This is an older node that doesn’t conform to the standard way of n8n processing each incoming item.
TBH this should be noted in the specific node documentation rather than in a generic looping page: Looping - n8n Documentation
Even I sometimes forget and have to go digging for the generic page to make sure