Imatest measures MTF using a variety of techniques (see table below), each of which gives different results in imaging systems because image processing depends on local scene content that is rarely constant throughout an image. Sharpening (high frequency boost) tends to be maximum near high-contrast features, while noise reduction (high frequency cut, which can obscure fine texture) tends to be maximum in their absence. For this reason, MTF measurements can be very different with different test charts.
In principle, MTF measurements should be the same when no nonuniform or nonlinear image processing is applied (e.g., when the image is demosaiced with dcraw with no sharpening and noise reduction). However, this situation never happens in practice because demosaicing, which is present in all cameras that use Color Filter Arrays (CFAs), involves some nonlinear processing. Noise, for example, is handled differently from edges.