It really depends on what you want to do;
1. If you are looking for trace levels of HMF, then you need optimum sensitivity. If you don't have a diode array detector, run the sample in a solvent resembling your HPLC solvent (assuming it is not a gradient) on an external UV spectrometer and use the peak maximum. If you don't have an external UV spectrometer, run iterative runs using the same sample injection size and different nm settings on the HPLC detector and use the nm with the highest response.
2. If you are looking for a crude purity determination (for example after a lab or process reaction), then you are likely assuming that the impurities have an identical response to your expected compound (which is a bad assumption if some impurities are aliphatic, for example). The difference between measuring at 280 and 284 nm is inconsequential.
3. For optimum analytical rigor (for example in the pharaceutical business), all of the unknowns in a process are identified with mass spec, characterized for their molar absorptivity coefficient at a particular wavelength and then quantified using accurate response factors for each of the known impurities.