Ummm... I don't see the problem. Your market value conclusion for the subject is based on a market exposure time of so many days, which says in other words that for a property similar to your subject, if it were placed on the market for the number of days you give for exposure time, starting that number of days before the effective date of the appraisal, it would likely sell for a price most probably equal to your value conclusion. So, in other words, you are indicating that your concluded market value is also valid for other homes in similar condition, sold in the same time frame.
Now, of course, there may not have been any other similar home on the market during that time frame (e.g., eff_date_of_appraisal - exposure_time_in_days to eff_date_of_appraisal), so you are inferring this value and exposure time from surrounding data. Likely, such inferences of exposure time are rough estimates; however, you could add exposure time for all sales to your data set and regress on that as well, to derive a formula for the impact of exposure time on net sale price. Using earth, you might want to expect the model will have additional interactions - but the problem is that it will have major interactions with condition and appeal, which are not quantized in your data. Note that if you regress on DOM, you, in effect, obtain an indirect and approximate measure of market appeal. The problem with using DOM is that it has a lot of "noise," e.g., DOM is impacted by real estate agent motivation, competence, and general market condition changes over time (that is to say the impact on net sale price of an exposure time of say 90 days depends on the date of sale).
So if running MARS on a data set with DOM as a variable, we would expect to see an interaction between DOM (i.e. exposure time) and Date-of-Sale (e.g Sale Date), effectively making the variable DOM useless as a measure of Appeal (e.g. condition, quality, **, design, ...). Although, as appraisers, we know that a DOM that very clearly doesn't fit the pattern is a sign that something is amiss.
Note: Usually, only MARS can make sense of the tangled mess of inter-relationships.