"W" expansion followed by "lgx" reduction doesn't fully reduce

[mmaiers-nmdp]

ard.redux_gl("DRB1*14:01/DRB1*14:54", "lgx")
'DRB1*14:01'
>>> ard.redux_gl(ard.redux_gl("DRB1*14:01/DRB1*14:54", "W"), "lgx")
'DRB1*14:01/DRB1*14:54'
>>> ard.redux_gl("DRB1*14:01/DRB1*14:54", "W")
'DRB1*14:01:01/DRB1*14:01:02/DRB1*14:01:03/DRB1*14:01:04/DRB1*14:01:05/DRB1*14:54:01:01/DRB1*14:54:01:02/DRB1*14:54:01:03/DRB1*14:54:01:04/DRB1*14:54:01:05/DRB1*14:54:01:06/DRB1*14:54:01:07/DRB1*14:54:01:08/DRB1*14:54:01:09/DRB1*14:54:01:10/DRB1*14:54:01:11/DRB1*14:54:01:12/DRB1*14:54:02/DRB1*14:54:03/DRB1*14:54:04/DRB1*14:54:05/DRB1*14:54:06/DRB1*14:54:07/DRB1*14:54:08/DRB1*14:54:09/DRB1*14:54:10'
>>> ard.redux_gl(ard.redux_gl(ard.redux_gl("DRB1*14:01/DRB1*14:54", "W"), "lgx"), "lgx")
'DRB1*14:01'

W expansion followed by lgx reduction should reduce all the way down but it apparently does not.
The brute force workaround is to call redux_gl again but this is less than ideal