The concept of "most probable price" in the market can be readily measured after the fact by what the market participants actually did as opposed to what some outside analysis concludes they should have done.
Well, there you go. That is absolutely incorrect. Certainly is also lacking in sufficient detail.
At the time the contract is signed, the "most probable price" is, in reality, what statisticians call the "expected sale price" - and it would be based on what market participants in the market area did prior to the actual sale transaction. That expected value would be your only objective evidence of "most probable price."
I will make a point here: It is taboo to consider any market sales transactions after the subject sales transaction. With MARS and similar techniques, we put a lot of effort into creating robust models to handle new transactions that have never been seen before. We do this through the partitioning of data sets and cross-validation. The model created, then, is generalized enough to allow it to do a satisfactory job of estimating the probable price for a property it has not yet seen.
So, to your original point, we should be able to see all relevant market sales transactions that go into estimating "the most probable price" before the contract is signed. Yes, they may not show up in the recorder's office until weeks later, and of course, some sales will be pending, and the contract price will be hidden for some time after closing. If the sales price for a transaction closing just before the subject's sales contract is signed, is hidden from view, then one might argue it should not be considered. However, note that when a buyer scours the market for a particular kind of property that suits his needs, he can often pick up this information informally through rejected offers or agent remarks. So, I would say if a comparable sales contract was signed before the subject sales contract date, then it should be used, as long as it is fairly similar to the subject or is something the model can adjust to.