I wonder how FDR data on control surface positions vs. flight crew inputs from this "landing" would compare to simulator control surface positions vs. inputs in a replicated scenario. If they vary significantly then the crew had zero or negative training on how to conduct a landing flare in direct law.
There's a reason the data for our (A320 drivers) new EEC training had to be derived from actual flight testing. There simply wasn't adequate programming to replicate full stall and upset conditions prior to the data collection flights. There may be a similar dynamic here.