Been looking at that ADDS thing a bit more. There's something seriously wrong with that algorithm. The interpretation is done correctly, as is the presentation. But then the wrong bits of TAF are associated with the presentation bits. Everything is well (excellent, actually) up to the PROB30 bit.
Well spotted. I think the key issue is that "PROB30 TEMPO" is fairly rare in US TAFs. If you speak SQL, here's the difference between the US and UK interpretations of the WMO's "TAFs for Dummies".
avbrief=# select count(*) from taf where icao ~~ 'K%' and text ~ 'PROB30';
count
-------
349
(1 row)
avbrief=# select count(*) from taf where icao ~~ 'K%' and text ~ 'PROB30 TEMPO';
count
-------
0
(1 row)
avbrief=# select count(*) from taf where icao ~~ 'EG%' and text ~ 'PROB30';
count
-------
205
(1 row)
avbrief=# select count(*) from taf where icao ~~ 'EG%' and text ~ 'PROB30 TEMPO';
count
-------
200
(1 row)
Looking at some of the examples, TEMPO seems to be implied by PROB30 in some cases.
Thus my guess is that the algorithm was not tested on cases where two change words like "PROB30 TEMPO" appear consecutively.