get_extended_trials fails for some behavior only sessions

For some behavior only sessions, running get_extended_trials() provokes a somewhat inscrutable error from within trials_processing.py.

For example:

>>> from allensdk.brain_observatory.behavior.behavior_session import BehaviorSession
>>> session = BehaviorSession.from_lims(962822728)
>>> df = session.api.get_extended_trials()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/allen/aibs/technology/danielsf/miniconda3/envs/allensdk_dev2/lib/python3.8/site-packages/allensdk-2.7.0-py3.8.egg/allensdk/brain_observatory/behavior/session_apis/data_transforms/behavior_data_transforms.py", line 261, in get_extended_trials
    return get_extended_trials(data)
  File "/allen/aibs/technology/danielsf/miniconda3/envs/allensdk_dev2/lib/python3.8/site-packages/allensdk-2.7.0-py3.8.egg/allensdk/brain_observatory/behavior/trials_processing.py", line 964, in get_extended_trials
    return create_extended_trials(trials=get_trials_v0(data, time),
  File "/allen/aibs/technology/danielsf/miniconda3/envs/allensdk_dev2/lib/python3.8/site-packages/allensdk-2.7.0-py3.8.egg/allensdk/brain_observatory/behavior/trials_processing.py", line 754, in get_trials_v0
    from_group, from_name, to_group, to_name = get_image_info_from_trial(trial_log, ti)
  File "/allen/aibs/technology/danielsf/miniconda3/envs/allensdk_dev2/lib/python3.8/site-packages/allensdk-2.7.0-py3.8.egg/allensdk/brain_observatory/behavior/trials_processing.py", line 696, in get_image_info_from_trial
    _, _, prev_group, prev_name = get_image_info_from_trial(trial_log, ti - 1)
  File "/allen/aibs/technology/danielsf/miniconda3/envs/allensdk_dev2/lib/python3.8/site-packages/allensdk-2.7.0-py3.8.egg/allensdk/brain_observatory/behavior/trials_processing.py", line 690, in get_image_info_from_trial
    raise RuntimeError('Should not have been possible')
RuntimeError: Should not have been possible

Here is a random sample of sessions that cause the error

[962822728, 1064436629, 1059690999, 877218464, 827655347, 896543686, 1034920219, 853805860, 1047709328, 1042765331, 1078567940]

Here is a random sample of sessions that do not

[818782934, 810303404, 882415298, 1010794220, 1026342864, 1007849594, 861788094, 778884020, 910859641, 814269306]

I have not put any effort into interrogating what makes these sets of sessions different.