I, like you, have heard this statement made by several people who are generally extremely reliable about such things...
but I've not yet seen the publication-- assuming that one exists at this point. I'm guessing, based on past history, that this may be a pre-publication finding from an in-progress study, or one recently wrapped up.
There must be something to it, though-- because I've heard the estimate 15-20% from one person, and "about one in five" from another. Your estimate is too close to either of those to be pure coincidence.
This is also something that I've heard is a concern (? and maybe more than a concern-- maybe "observed side effect" even) in oral immunotherapy trials, though the numbers are pretty small in those trials, so it's not yet clear to me that they really know incidence very well.