They'd get too many false-positives. LOTS of people use a cell phone as their only phone. And for an elderly shut-in on fixed income, who "doesn't get out much", that phone isn't going to move very much (but it will be plugged in via "wall-wart" so that it can stay ON all the time, in case a call comes in.)
There would be a pattern to this, that could be observed at the central cell sites. A phone in the tracfone category would appear on the local roaming database (it has to check in with the cell periodically to let the cell know it is there to get calls), and would stay there day after day, never ever making or getting calls.
I know that there are people, lots of 'em, who fit that profile -- honest, law-abiding people. Trying to pick the tangoes out of that mix would be like trying to pick flyspecks out of a peppershaker.
Now, in theory, they could try to build a database of "residential" and "non-residential" locations, and then cross-reference low-activity "registered" phones to that database, but by the time that got that database constructed, and the queries coded and debugged (including the interface code)... well, it'd be far into the "too late to do any good" timeframe.
A tracphone doing this, sitting stationary at a major public site. would be a possible indication of trouble.
Oh and your "database" is as close as, say, Microsoft Streets and Trips. A terrorist is unlikely to set a bomb at Granny's house, but may want to do so at an overpass.