>On average they reject 98%. Mensa's cut off is the top 2% of IQs
These two statements are not congruent unless the distribution of IQs in people applying to Mensa is equal to the ditribution of IQs in the general population. I doubt that to be the case.
If I understand it correctly, they claim that you must be in what is thought to be the top 2%, in other words your IQ must be over 130-something.
But this does not mean that they reject 98% of the people that take the tests. People interested in taking that test must already be, I suppose, in the top 10% of the IQ distribution (125+?), or so they think. If this is true they they do reject 98% of the applicants but only 80%. Yet I suspect they reject even less.
Any source or third-party evaluation of their acceptance rate?
AFAIK, tests are prepared by some third party organizations (psychology departments?), scaled on representative sample of the population and are just bought by Mensa(and changed every couple of years). You should be aware that in different countries test with different SD (15, 16 and 25) are used. I heard from one guy who was responsible for testing in my country that about 25% of people taking the test get accepted
