They aimed to recruit 80 interviewees. 20 L, 20 G, 20 B, 20 T. That in itself is a bit iffy - there are for instance almost certainly more bi people than lesbian and gay put together - but let's accept it as a way of getting a broad L+G+B+T cross-section for now.
They recruited 20 lesbians, 22 gay men, 21 transgender people, and ...13 bisexuals. Apparently, we were really hard to find. So they gave up looking?
And then they interviewed this group of 76 people (apparently no-one was both B and T, or what have you) and picked the most pertinent quotes from the panel to intersperse through the report, as you do when a long wordy publication needs a bit of colour. Quotes were anonymous but tagged [G] for a quote from a gay man, and so on. I thought there seemed to be a bit of a dearth of [B] so counted them up, as I had a long train journey to pass.
Lesbian: 20 people, 31 quotes
Gay: 22 people, 32 quotes
Transgender: 21 people, 27 quotes
Bisexual: 13 people, 12 quotes
Gosh. L, G or T, get interviewed and we'll use on average one and a half snippets of what you contributed. Bis? One quote each, form an orderly soundbite queue please.
So: are bis boring, is this within normal statistical distribution, or is this a somewhat suspicious bit of subediting from GEO?