The default is set that way because Doug Ward, the author of the first version of 3dDeconvolve, felt that this was a way to treat all the lags equally with respect to the data points that they could possibly be influenced by.
However, I do not recommend using the minlag and maxlag method of deconvolution any more. You should use the -stim_times method, which is simpler to use under most circumstances, and more flexible.