I vote no as well.
Indeed, in a CE presentation I give, I argue that the lack of data to populate the 1004MC provides reason and supports the decision for going (a) further back in time, (b) farther out in distance, or (c) using properties that would not normally be categorized as comparable to the subject.
So I'd populate the 1004MC with the 2 sales and zero listings that the comparable search captures (of course, like you, I clearly state what the search criteria is... in fact, I'd argue this is required information on the form). Then, I state that it is evident from the 1004MC analysis that ideal sales to use as comparables do not exist. Therefore, I did blah, blah, blah; the result is the 4-sales in the grid which represent the best comparables to consider and analyze for this assignment.