The study itself explicitly states that they don't have the data to reach any conclusions about the relative safety of splitting vs not.
What the study shows is that riders who were lane-splitting at the time of an accident sustain fewer injuries than those who weren't. It mentions some differences between the two groups -- lane-splitters are more likely to be commuting and have better safety gear -- but glosses over how that impacts the injury rates.
I don't have another study to cite but motorcyclists who commute are likely a much less injured group than recreational riders. For a host of reasons, from not engaging in as much risky behavior -- who wants to die on the way to work? -- to being more skilled on average due to experience.
In my experience, lane splitting is much safer. It has saved me from violent rear-end collisions on several occasions.
I'll also go as far as to say that commuting is certainly more dangerous than most recreational riding (the exception being recreational riding when the drunks are out, e.g. midnight on the weekend) due to traffic density. The text-and-drive crowd is also more active (both on their phones and on the road) during rush hour, perhaps because driving inattentively seems like a safer practice at lower speeds.
Commuting being a seemingly more dangerous activity doesn't preclude commuters from being a lower-risk-of-injury group. Age could be another factor -- older riders are more likely to be injured in motorcycle accidents, and might be less likely to commute or lane-split.
The study is crap. It lacks the data to control for other factors and draw meaningful conclusions about much of anything.
I don't see where in that study it shows what you claim. They specifically say:
> This study is not without limitations. The primary limitation is our lack of exposure data. To estimate how the risk of being involved in a collision changes when motorcyclists chose to lane- split, we would require information on both the lane-splitting and non-lane-splitting riding that is done by some identifiable sample of motorcyclists. The collection of these data is fraught with problems, and the current study did not attempt to collect such data. The current data set cannot be used to compare the collision risks for lane-splitting or non-lane-splitting riders.
https://news.berkeley.edu/2015/05/29/motorcycle-lanesplittin...