Most human judgment includes comparisons of a target object with one or more reference objects. Comparisons are known to sometimes lead to judgment biases. We perceive the moon as larger the closer it gets to the horizon. We feel water at room temperature is warmer if our hand was submerged in cold water just before. We estimate things are smaller than they really are when we compare them with smaller objects and larger than they really are when we compare them with larger objects. Software development effort estimation is also frequently based on comparing the effort of one task relative to that of another. We present empirical results that show how relative estimation may result in biased assessments of similarity and over-optimistic effort estimates. We observe that tasks tend to be assessed as more similar than they are in reality when compared with each other, that the similarity of two tasks depends on the direction of the comparison and that it matters whether the comparison is based on difference in work-hours or as the ratio. We also observe that the use of the Fibonacci-scale for agile estimation sometimes has unwanted side-effects which may lead to under-estimation of tasks. We use these observations and other evidence to suggest ways of improving the accuracy of relative estimation as it is done in, for example, agile projects.