Look at integrated MOSFET construction. Can't you see
the N, P, N structure (or P, N, P for PMOS)? There you
are.
Suppressing the BJT is job #1 for device design / process
development / layout groundrules definition. But every
decade somebody rediscovers some way to put the BJT
to use. And then generally gives up, because it's not a
very good transistor.
Lateral BJTs have only one good point, which is a
symmetric high blocking voltage. You saw them used for
this, specifically, in standard linear ICs for decades.
A lateral PNP was (1) a freebie, using the NPNs' implants
and (2) a way to get 30V input differential voltage in
a 30V process.
Lateral BJTs otherwise svck pretty badly especially
when it comes to low current beta rolloff (due to a
high base surface area and that the collector current
has to run along the underside of the base oxide
interface which nobody really cared to optimize for
traps / surface recombination velocity (especially
not CMOS process developers, for whom any kind
of recombination is generally a bonus).
"Love the one you're with" is about it. She's dumb
and ugly, but PhD supermodels rarely show up around
closing time.