It is the same if I have just one single row as result of a query. E.g. a query that returns a single integer, ran in the Data Explorer panel in the admin backend it takes 9 ms, ran from the outside against the Data Explorer API via curl it takes 200 ms.
I would very much like to avoid that, I use Data Explorer because it was suggested on these forums as a nice way to get information from the DB and a well working API for that.
My query uses them for JOINs but not in the result set. The SELECT only has one column and that is a newly calculated one with a arbitrary name.
Assuming that those are the floating point numbers 3 places right of the HTTP status code (e.g. ... 200 642 "-" **0.107 0.108** "system" ..., highlighted with stars by me) they are about 10 times the duration of the query time reported in the backend and pretty much exactly half the time it takes to curl from my client.