Further to the various comments (especially noting latexman who sensibly refers back to API STD 520 Part II), it's not uncommon for relief calculations to overlook the importance of "total nonrecoverable pressure loss" when assessing the 3% inlet loss criteria. Non-simple piping configurations can give rise to pressure recovery, so it's important to assess losses accurately that are consistent with the criteria definition.
The Smith, Burgess and Powers paper that latexman has uploaded in Post #3 is a worthwhile read but it's effectively superseded by the more complex methods cited in API STD 520 Part II. I've personally used the method proposed by Melham, "Analysis of PRV Stability in Relief Systems, Part II - Screening" to do a PSV force balance assessment for brownfield work in cases where the 3% rule was violated. The Melham paper provides a clear method that is not too difficult to put into spreadsheet form.
The paper (and its other parts) can be sourced from the ioMosaic website.
I can see a future where the 3% rule will be consigned to history and a force balance check will become standard practice. Until then, if you're less than 3% on the upstream side of the PSV, the design is compliant with the current standard.