Possible bug in PhenomXPHM start time

Alex Nitz would like to report a bug with PhenomXPHM but is able to make an issue here so I'm pasting his comments below.

Pinging PhenomX devs: @cecilio.garcia-quiros, @geraint.pratten, @sascha.husa and reviewer @maria.haney

Is this a bug?

""" I would like to report a bug in IMRPhenomXPHM. It isn't setting the end time for both polarizations correctly. I just generated a waveform and the plus polarization is at -32 seconds, and one is at 0. They must by convention bet set to the same value, and only the latter seems correct.

 [4]: hp, hc = pycbc.waveform.get_fd_waveform(approximant="IMRPhenomXPHM", mass1=80, mass2=64, f_lower=30, delta_f=1.0/32)                                  
In [5]: hp.start_time                                                                                                                                         
Out[5]: LIGOTimeGPS(-32, 0)
In [6]: hc.start_time                                                                                                                                         
Out[6]: LIGOTimeGPS(0, 0)

It looks like it's setting hp to -duration ( corresponding to the delta_f). If time were really cyclic, this would be the same point, but downstream codes expect this to be a meaningful value to determine where the reference peak is for the time series equivelant. This will either break outright (as I found) or be interpreted as one polarization occuring 32 seconds before the other in other cases. (edited) """