raw.crop failure with multi-file CTF

I'm unsure about the exact meaning of raw.first_samp, but I think there might be an error when cropping multi-file CTF recordings.

I'm working with ~1h long CTF recordings that I crop into blocks to preprocess. Unfortunately the CTF recordings is big (>10GB), so hard to share. The following demonstrates the problem:

>>> import mne
>>> raw = mne.io.read_raw_ctf(filename)
ds directory : /home/gortega/megdata/S1-5_Attractor_20161022_01.ds
    res4 data read.
    hc data read.
    Separate EEG position data file read.
    Quaternion matching (desired vs. transformed):
      -1.14   80.92    0.00 mm <->   -1.14   80.92    0.00 mm (orig :  -53.00   56.58 -262.06 mm) diff =    0.000 mm
       1.14  -80.92    0.00 mm <->    1.14  -80.92    0.00 mm (orig :   48.86  -69.11 -267.17 mm) diff =    0.000 mm
     100.30    0.00    0.00 mm <->  100.30    0.00    0.00 mm (orig :   75.61   54.06 -244.96 mm) diff =    0.000 mm
    Coordinate transformations established.
    Polhemus data for 3 HPI coils added
    Device coordinate locations for 3 HPI coils added
    64 EEG electrode locations assigned to channel info.
    64 EEG locations added to Polhemus data.
    Measurement info composed.
Finding samples for /home/gortega/megdata/S1-5_Attractor_20161022_01.ds/S1-5_Attractor_20161022_01.meg4: 
    System clock channel is available, checking which samples are valid.
    1104 x 1200 = 1324800 samples from 405 chs
Finding samples for /home/gortega/megdata/S1-5_Attractor_20161022_01.ds/S1-5_Attractor_20161022_01.1_meg4: 
    System clock channel is available, checking which samples are valid.
    1104 x 1200 = 1324800 samples from 405 chs
Finding samples for /home/gortega/megdata/S1-5_Attractor_20161022_01.ds/S1-5_Attractor_20161022_01.2_meg4: 
    System clock channel is available, checking which samples are valid.
    1104 x 1200 = 1324800 samples from 405 chs
Finding samples for /home/gortega/megdata/S1-5_Attractor_20161022_01.ds/S1-5_Attractor_20161022_01.3_meg4: 
    System clock channel is available, checking which samples are valid.
    1104 x 1200 = 1324800 samples from 405 chs
Finding samples for /home/gortega/megdata/S1-5_Attractor_20161022_01.ds/S1-5_Attractor_20161022_01.4_meg4: 
    System clock channel is available, checking which samples are valid.
    726 x 1200 = 871677 samples from 405 chs
    723 samples omitted at the end
Current compensation grade : 0

There are 5 meg4 files for this dataset and croping beyond the first meg4 file doesn't work:

# Crop within first meg4 file:
raw.copy().crop((100+1)/1200, (100+10)/1200).first_samp
101 # This is OK!
# Crop in 2nd meg4 file:
>>> raw.copy().crop((1324800+1)/1200, (1324800+10)/1200).first_samp
1 # I believe this should be 1324800+1

I think the source of the problem is that raw._first_samps and raw._last_samps are not set correctly:

>>> raw._first_samps
array([0, 0, 0, 0, 0])
>>> raw._last_samps
array([1324799, 1324799, 1324799, 1324799,  871676])

Fixing them like so solves the problem:

>>> raw._first_samps = np.cumsum(raw._raw_lengths) - raw._raw_lengths[0]
>>> raw._last_samps = np.cumsum(raw._last_samps)
>>>  raw.copy().crop((1324800+1)/1200, (1324800+10)/1200).first_samp
1324801 # OK!

But this leads to problems for me down the line.