Not enough information. What is your assay? On what materials? Experimental error/variation in assays in geochemical analysis (for example) is typically 98 - 102%.

We had this problem when performing the accuracy of the Assay / UC / Blend method. At first we thought that the problem was due to the hypromellose, but then, we prepared a sample only with standard, but we always recovered 102%. When we decrease the amount of methanol we reach 101%. There is a clear influence of methanol, but I do not know why.

It looks like you have a binding problem of the API to the glass. The standard is probably not in methanol solution. The API in the drug product will have less binding with increased methanol and thus the increased recovery. Recommend low binding vials and use of the standards in the same solvent as the drug product.