The EPA says consider fixing in that band, since no level is risk-free. The sensible next step is a long-term test across a heating season, plus an honest look at who uses the lowest level and whether anyone smokes.
More detail
A borderline number deserves better data before it deserves money. Short tests swing with weather and season, and a July 3.1 can be a January 5.5, so a long-term alpha-track detector riding from fall to spring produces the average worth deciding on. While it runs, weigh the household factors the bare number omits. A lower-level bedroom or a daily-use basement office multiplies hours of exposure; a smoker in the house multiplies the risk per hour; a family planning twenty years in the house accumulates more than one planning two. Any of those can justify fixing a 3 that a childless couple upstairs might reasonably watch instead. The economics stay gentle either way: mitigation pricing does not care whether the starting number was 3 or 13, so waiting for better data costs nothing but the detector. The band-by-band walk-through is in the borderline result decision guide, and the diagnosis page sorts your situation in two minutes.