According to the wikiislam link, the sunnipath link was created in October 5, 2007. The link is dead and the earliest Archive of it was captured on 26 October, 2007 by Wayback Machine. This captured snapshot says nothing about the claim and talks about a study about Muslim divorce rates:
One of the oft quoted statistics about divorce rates amongst Muslims in North America comes from an early study done by Dr Ba-yunus.
But the divorce studies methodology is disputed by other other researchers.
In the limited amount of research available on marriage in the
American Muslim community, there is conflicting evidence about the
frequency of divorce. A single study estimates a nationwide American
Muslim divorce rate of 32.33% (Ba-Yunus, 2000, 2007), a high number
and one that is contested (Alshugairi, 2010; Leonard, 2003) because of
the methodology of the study. Ba-Yunus derived his statistic by
examining 10 years of official marriage and divorce records in five
American states and one Canadian province selected for their large
Muslim populations. He divided the average number of Muslim divorces
per state by the average number of Muslim marriages per state, and
then averaged the states’ mean Muslim divorce rates to yield a
“nationwide” statistic of 32.33% (Ba-Yunus, 2007). This statistic
should be considered with caution, not only because of the lack of
information about the sample and the non-random selection of states,
but also because of the difficulty of obtaining accurate figures for
Muslim marriages and divorces. Not all Muslims will document their
marriages (or divorces) with their state. Some prefer religious
ceremonies to legal ones (Macfarlane, 2012), and others were married
in a country outside the United States, so that the only American
legal record is of their divorce (Ba-Yunus, 2000).
In contrast to Ba-Yunus’ divorce rate of 32.33%, Alshugairi (2010)
reported a much-lower divorce rate of 21.3% for a sample of 751
Californian Muslims. This sample statistic closely matches the
above-mentioned ten year divorce rate of Asian American women – an
important point, since approximately one-third of American Muslims are
South Asian (Bukhari, 2003). Yet it is Ba-Yunus’ higher statistic that
appears in several mainstream American Muslim publications (e.g.
Ghayyur, 2010; Kholoki, 2007; Siddiqui, 2009), despite the lack of
consensus over its accuracy. Ba-Yunus’ statistic has caused and
reflected considerable unrest. Further empirical investigation of
American Muslim marriages may help to allay concerns or point toward
specific problems that may be investigated in more detail.
American Muslim Marital Quality: A Preliminary Investigation
According to the video G. Bach linked to in comments, Ilyas Ba-Yunus said (reportedly) [at 9.15]:
According to our estimation ... we privately found that out of 100 people who accept Islam, 75% of them give up on Islam
This is not a study. This is what he reportedly said and his words were recorded. The study doesn't exist. Also, the site wikiislam has huge number of statistics that are outright false and not staticstics at all. For example:
The third link refers to this (webcitation.org link. CNN deleted it from their own site) CNN article which says:
Conversion will play relatively little part in the increase, the report anticipates. It says little data is available on conversion, but what little there is suggests Islam loses as many adherents via conversion as it gains.
No, that's not true, that's fake News. CNN commentator made up that lie. According to The Future of World Religions: Population Growth Projections, 2010-2050, Islam is expected to gain more than 3.2 million followers due to religious switching, the largest for any religion.