Pavel Suderevsky <psuderevsky(at)gmail(dot)com> writes:
> During performance tests it was mentioned that pgbench common results are
> being different from what pg_stat_statements provides.
Well, they're not measuring even approximately the same thing, so I'm
not sure why that surprises you. pg_stat_statements only measures
the executor runtime, omitting parse/plan times as well as network
transmission times. pgbench is reporting the overall time as seen
from the client side.
regards, tom lane