This thread is interesting and thought provoking. Ever since I've learned that this weighting issue is the Achilles heel of appraisers in the review/underwriting, I've gone with the straight line average non weighted approach myself, in most instances.
You know I had a thorough post, but baby pulled the power cord. Bummer man. So here it is again in a nutshell.
Your adjusted end values in the grid already 'carry the weight'. Nothing good comes from keeping net/gross down and applying weighting to already adjusted valuation figures instead. The grid is there to build them up and break them down. Variances in housing is adjusted back to align with the subject, so unless you've got an a-typical comp that does not fit into your static adjustment scheme, you don't need to weight comps who's adjustments already carry the weight of market differences.
In theory, the above has some validity, but in reality every adjustment inherently comprises an element of judgment, therefore, the greater the adjustments overall (on a gross basis), the less reliable a comp becomes (otherwise, there would be no difference between 'sales' and 'comps' since theoretically anything could be reliably and equally adjusted to an exact substitute for the subject regardless of its differences). The cummulative nature of the judgment and subjectivity in the adjustment process logically dictates that those comps requiring the most adjustments generally result in the least reliable indication of value. For this reason, if one is to apply a weighting (or emphasis) on certain comps after adjustments, the most logical means to do so would be to base it off of total gross adjustments.