I found in an article that if crystallization area below 30 nm DSC can not find the crystallization in solid dispersion, but DSC is a very promising tool for characterizing the polymer blends.
If you are looking for spatial distribution of the crystallization area, AFM is more accurate, if you are looking for quantity ratios, the DSC is more accurate. Because AFM traces the surface only and upon mapping the variation of the response over the surface it can give good map of the crystallization areas, unlike DSC that gives results for bulk quantities. So, it depends on your purpose.