Apologies, my post seems to have lost the previous thread from January where I originally raised this issue (see https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ACB-AKI-ALGORITHM;c1653b77.1501 ). 
This scenario occurs if the results concerned are entered close to each other in time (within a few minutes). The algorithm appears to be unable to avoid using creatinines dated after the result currently being calculated.
This may have been a poor example for "real world" consideration - no result would be reported a week after sampling! - but this could easily occur for samples taken within a day or so of each other.