If both FTS and ID base search are supported for a specified or implied property, then exact matching will be used where possible.
For example, the following query will match running but will not be tokenized. If you are using stemming it might not match anything.
For the cm:name field, which is in the index as both tokenized and untokenized, it will use the untokenized field. For example, =part will only match the exact term "part". If you use =part* it will match additional terms, like "partners". If there is no untokenized field in the index, it will fall back to use the tokenized field, and then, with stemming/plurals, it would match.