I was looking at Malew c.1765 marriages today - if you went on extracted records only 45% or so would be found though those had a 95% or better probability of being correct.
Patron submitted entries overwrote the rest - here the odds are that the entry was incorrect - wrong year, misreading of publication of banns as marriage date, misreading of name (one batch was consistently wrong being a year out each time) - Maughold is much better only 5-8% or so error rate usually in misreading name