The problem you are facing is due to the artificial delay added in the time domain simulation for the waveguide element. This problem is fixed and will be available in the next release (targeted at this May, so will be available very soon). Corresponding documentation will also be online together with the release, more explanation will be in there.
The primitive ring modulator element has build in delay compensation and build in FIR filter while the compound built by couplers and waveguides does not, so the time domain simulation results are different.
The following figures are the waveform measured by the oscilloscope for the primitive ring modulator, compound ring modulator without delay compensation and with delay compensation, respectively, with same settings (2 Gbits/s, 64 bits, 4096 samples/bit). You can see the result of the compound with delay compensation is very similar to the result of the primitive one.
For the ring modulators (generally for all resonance structures), the sampling rate should be large enough so that the time domain simulation matches the frequency domain characterization (impulse response matches the scattering analysis). Another trick, add a low-pass filter after the PD will help to eliminate the over/under-shoot in the waveform. The modified file may help a little bit with a more reasonable global setting: single_ring_5um_modified.icp (1.6 MB)
For the FWHM bandwith, the Q-factor measured at 193.333 THz is about 9664.73 (~10000). So the 3dB BW should be ~20 GHz, which is correct according to the gain curve measured by the ONA:
I hope this could help to some extent The new version of IC will be available soon to solve this problem.