Whether or not, and how you setup to record the MIDI data generated by the MOXF arpeggiators depends in large part on the KEY MODE. Briefly, the Key Mode determines whether or not the actual keys you press are documented in the sequencer, or whether the notes output by the arpeggiators are recorded to the sequencer.

In general, the MOXF sequencer can record one or the other. This conundrum is caused by the keyboard, the arpeggiator and the synth Part all share the same channel number. Both the keyboard and the arpeggiator are transmitting to the tone engine. In order for you to send a message to the arpeggiator and the arpeggiator to trigger the tone engine, the channels must match.
This is a problem because a Midi track would simply record all midi messages, so the default is to record one or the other. Since typically, you want to filter the trigger notes and record the notes output from the arpeggio phrase that is the default.

Divide and Conquer
The MOXF can record all of this in real time, but it requires you rebuild your velocity zoned Voice as a Performance. Giving those Elements that sound in direct response to trigger notes their own PART. And those Elements that are going to generate Midi notes from an arp phrase their own PART. Then using Direct Performance Record each of your different articulations will be faithfully recorded.

This means making separate Voices from your original single Voice, so that they can be separate components of your new Performance.

Direct Performance Record allows you to play multiple Parts simultaneously, and when routed to the sequencer - each Part is separated to a separate Midi channel, and each Part of the Performance is copied to the Song/Pattern Mixing you target, such that each will align with its separated Midi data.
By separating the Midi channel of the "direct" notes, from those "sorted" by the arpeggiators, and recording the data to separate tracks, you can work around the conundrum.

Make one Voice with the KEY MODE changed to "sort" from the "sort+direct" setting.
Make a copy of this Voice to an adjacent Voice location, and set the KEY MODE = "direct".

Place these two Voices in a Performance (being sure to use the Parameter with Voice feature to automatically assign the arpeggiators to the Parts) and then use direct Performance Record to record to the sequencer.