For this passage, the 1852 LDS edition changed the singular tradition to the plural traditions, probably because of the following plural were (thus eliminating the subject-verb disagreement found in the earliest text, “and of the tradition ... which were not correct”). Oliver Cowdery was the scribe here in 𝓟 and probably also in 𝓞 (extant fragments of 𝓞 in nearby Alma 20 and Alma 22 are in Oliver’s hand). And there is considerable manuscript evidence that Oliver frequently mixed up the number for tradition(s); for six examples, see under Mosiah 1:5. Thus here in Alma 21:17, one could interpret this occurrence of tradition in 𝓟 as an error for traditions, especially since the verb in the following relative clause is in the plural (“which were not correct”).
Yet we also find in the original text that the number agreement for the verb in the relative clause is often determined by the number of the nearest preceding noun, as in 1 Nephi 13:23: “a record of the Jews which contain the covenants of the Lord” (see the discussion under that passage). Therefore, the critical text will accept the earliest extant reading here in Alma 21:17, “the tradition of their fathers which were not correct”, despite its problems with subject-verb agreement. Even so, we must still recognize that tradition could be an error for traditions.
Summary: Restore the singular tradition in Alma 21:17, despite the fact that the verb in the following relative clause is in the plural (“which were not correct”); such subject-verb disagreement derives, it would appear, from the nearer plural noun, fathers: “the tradition of their fathers which were not correct”.