The closer you are to something the more likely you are to hit it. That isn't hypothesis. It's basic physics.
Physics plays no part - unless there is a collision rather than passing very close. It is the softer human sciences that are in play here. Another hypothesis (they are easy to manufacture but dammed hard to prove/disprove):
See cyclist without helmet. Obviously an amateur tosser about to wobble everywhere. I will give him a wide berth.
If I am right (about his skills) the chance of a collision is x.
See cyclist with helmet. Obviously someone who cares about his safety and is unlikely to swerve unexpectedly to the right. I pass at a closer point but one with the same perceived chance of intersecting . So if I am right (there is a correlation between skill, care and wearing a helmet. then the chance of collision should not be higher than x. Indeed it may be lower.
Now one can argue (as did the original study) the wobble factor was actually the same (as it was the same rider) and hence the second perception (which was then wrong) would indicate (but no prove) a higher chance of collision.
Or perhaps ..
Real cyclists who choose/not choose helmets are usually different and may themselves behave differently. For example with a helmet the cyclist can feel 'safer' and hence take less care whereas the non-helmeted may feel more vulnerable and take care to wobble less when being overtaken.
In which case wearing a helmet is lethal. But we just don't know. It is dangerous to believe we - or physics - do.