All adjustments are derived from the market
What if the variables in question are not in the market. For example, your subject is the only property that has an X, a Y or a Z. Every time that happens, there will be no "pair" with which to make a paired sale. What if you subject is just way larger, smaller, older, newer - that is, just way out of the range - on measured variabes. There is nothing wrong with matched pairs, if you can find them. Good luck.
In my experience, most adjustments are just made up; and supported with some claim of market extraction, but you may never see it in a report.
My advice is to study "relative ranking." It is demonstrated in the AI text. Start by simply ranking sales as better, worse or roughly equal to subject. Determine where within the range subject's value is. I assume you are doing only form reports. You can put the ranking table in the report (and if it was a commercial report you wouldn't need anything else). Go to the line adjustment grid and make some small pro forma adjustments (say, for GLA, lot size, age-condition or extras) to show how the range of value closes around you conclusion. That is, use relative ranking to solve the problem, while using the grid as a "visual aid." Then as time goes by, you might gain confidence in using adjustments, or better ye, invest some time in learning regression analysis, which is by far the best way to get "adjustments" from the market.