With this information alone, no we cannot tell.

For example if we took random samples of the population and tested them for marijuana usage, what percentage would test positive?

Next, this study is only talking about marijuana testing, how many of the same group also tested positive for alcohol (or other impairing drugs). Lets make up fake numbers and say 60% of total fatalities had alcohol or other impairing drugs and the overlap between them and marijuana use was 80% then marijuana is rather insignificant.

We have to have all the details so we don't fall into a base rate fallacy.

Well, its the wrong universe of analysis to make that claim and there is no comparative measure of alcohol exposure in the same universe of analysis so it also fails to provide a basis for any alcohol/THC comparison, so, no?

It is entirely possible THC is more dangerous for vehicle safety than alcohol and that wouldn’t surprise me. But based on this study alone? No.

And likewise, what share of that 40% were killed by the other 60%? Fault was not evaluated here.

Yeah, drug use is also influenced by social and economic status, which also influences driving risks. People of lower socioeconomic status drive less safe cars on less safe roads for longer commutes. This is something valid to evaluate with a drug like THC which is detectable long after use. It would be nice to see the distribution of levels detected and not just the average.