Yes, and I measured it too, it has a fair bit of jitter:
The test was conducted as follows:
Tempo 120BPM
1 track playing short test pulse sample on 16th notes, 1 bar pattern
Record audio into computer from line out for 30 seconds.
Once recorded pick a section of 8 hits from within the 30 seconds
Truncate the audio file so that the first pulse is at the very start of the file, eg 0 samples from start.
In software set markers to 120BPM, then measure how far from each marker the pulse hits.
I tested synced to Octatrack, synced to Roland SBX1, and just 707 internal clock.
The number of samples deviation at 44.1khz is shown, a minus number indicates an early hit, a positive number indicates a late hit. For reference 100 samples is roughly 2ms, not easily heard in isolation, but very easily heard when synced to an accurate device with less jitter on percussive sounds especially. Also can be a problem when wanting to sample accurate loops etc.
I wrote to Roland asking for it to be improved, you should too
Edit: As @Oxenholme and @cogsy say sync to OT is acceptable. Strangely Roland SBX1 is known to put out good midi clock, yet 707 does not sync great to it.