Abduction, or inference to a best explanation, is a ubiquitous type of inference that is frequently used by humans in a wide range of tasks. However, many realistic domains have properties that make abduction computationally intractable (i.e., where the time to reach a solution increases exponentially with the number of possible explanations). We present a domain task analysis and performance evaluation of RedSoar, a plausible cognitive computational model of abduction, that accomplishes the antibody identification task in immunohematology. The task analysis reveals how a computationally intractable abductive problem, where one is seeking optimal solutions, can be reformulated to be a computationally tractable abductive problem, by seeking satisfactory rather then optimal solutions. From the satisfactory perspective, our evaluation framework of RedSoar's performance explores the computational benefits and costs of having directly available abstract hypothesis formation knowledge, and how a strong causal constraint between hypotheses and data reduces the combinatorial explosion of constructing a best explanation.