Update the second paragraph to the following text (noting that the underlined text is new content):
To ask an MPI to match a patient, clients use the "$match" operation, which accepts a patient resource which may be only partially complete. The data provided is interpreted as an MPI input and processed by an algorithm of some kind that uses the data to determine the most appropriate matches in the patient set. Note that different MPI matching algorithms have different required inputs. Consult with the vendor implementing the $match operation as to its specific behaviors.
The generic $match operation does not specify any particular algorithm, nor a minimum set of information that must be provided when asking for an MPI match operation to be performed, but many implementations will have a set of minimum information, which may be declared in their definition of the $match operation by specifying a profile on the resource parameter, indicating which properties are required in the search.
The patient resource submitted to the operation does not have to be complete, nor does it need to pass validation (i.e. mandatory fields don't need to be populated), but it does have to be a valid instance, as it is used as the reference data to match against.
Implementers of the $match algorithm should consider the relevance of returning inactive patients, particularly ones associated with patient merges.
E.g. If an inactive patient is "matched" and its merged target resource will be included, then inactive one may be excluded, however if a patient was just marked as inactive for other reasons, it could be included in the results.
(any specific MPI algorithm may or may not behave as in these examples)