The number of small radiation-induced DNA fragments can be heavily underestimated when determined from measurements of DNA mass fractions by gel electrophoresis, leading to a consequent underestimation of the initial DNA damage induction. In this study we reanalyzed the experimental results for DNA fragmentation and DNA double-strand break (DSB) yields in human fibroblasts irradiated with γ rays and nitrogen ion beams with linear energy transfer (LET) equal to 80, 125, 175 and 225 keV/μm, originally measured by Höglund et al. (Radiat Res 155, 818–825, 2001 and Int J Radiat Biol 76, 539–547, 2000). In that study the authors converted the measured distributions of fragment masses into DNA fragment distributions using mid-range values of the measured fragment length intervals, in particular they assumed fragments with lengths in the interval of 0–48 kbp had the mid-range value of 24 kbp. However, our recent detailed simulations with the Monte Carlo code PARTRAC, while reasonably in agreement with the mass distributions, indicate significantly increased yields of very short fragments by high-LET radiation, so that the actual average fragment lengths, in the interval 0–48 kbp, 2.4 kbp for 225 keV/μm nitrogen ions were much shorter than the assumed mid-range value of 24 kbp. When the measured distributions of fragment masses are converted into fragment distributions using the average fragment lengths calculated by PARTRAC, significantly higher yields of DSB related to short fragments were obtained and resulted in a constant relative biological effectiveness (RBE) for DSB induction yield of 2.3 for nitrogen ions at 125–225 keV/μm LET. The previously reported downward trend of the RBE values over this LET range for DSB induction appears to be an artifact of an inadequate average fragment length in the smallest interval.