For a failure or action level our rule of thumb (policy) would be to resample without taking any improvement measures. If the resample comes back clear then we would attribute it to contamination in some part of the sampling process. We have not looked into the frequency of endotoxin failures that clear on resample against those for CFU so unsure if one is more indicative of technologist contamination than the other.
The majority of our endotoxin measurements come in at <0.050EU/ml (current limit of tests we perform). I've noticed (so nothing scientific) that if the endotoxin results come back a fraction higher, say 0.056EU/ml, this is often accompanied by the accompanying CFU count reaching action level or failing.