I don´t think treating it as zero is correct, first you have to investigate why you get a negative value. Either you have a background problem (cleaning of the sample) due to the high concentration of sugars. Or the problem is with the precsion of the analysis, if you have a very large dispersion on the most concentrated points of your calibration curve the regresion analysis will have a very large effect on the lower concentration points where I assume is the conc of HMF.
I would like to thank everyone for valuable answers.
Dear Mario, yes you're right, after seeing your answer i read the following sentences from International Honey Commission (2009);
"At low HMF levels (about 5 mg/kg) the values obtained with this method (White) are comparable to those obtained with the HPLC method, but are lower than those obtained with the Winkler method (4). At higher HMF levels (20 and 40 mg/kg) the values with all three methods are not significantly different from each other"
In this case, i can say our honey has HMF < 5 mg /kg according to white method.