Well, even in a study about an easier thing such as the life expectancy of smokers versus non-smokers, it is impossible to entirely factor out other causes of death only indirectly related to smoking. Fictive examples:
-Smokers, more used to smoke, avoid cars' and trucks' smoke less, since it doesn't make them cough, and get more cancer from those;
-Smokers are people who tend to care less about risk and want immediate pleasure, so they would use drugs, drive too fast and expose themselves to danger more thus lowering their life expectancy.
In any case, since HRT does not bring significantly higher health risks for the vast majority of patients, compared to non-HRT, if it is prescribed by a competent specialist and appropriate follow-up is done (IMO, but I highly doubt I'm wrong), I still think that this statistic, if it is accurate, is 99 % attributable to:
-Suicides
-Murders
-Social integration problems (employment, less friends thus more depression, etc.)
-Self-medicating (wrong dose, illegal HRT not being the right product, people self-medicating oral HRT whilst they're at risk for liver problems and any competent doctor would've prescribed transdermal...)
-Other co-morbidity related issues such as depression, drugs, prostitution to pay for transition, transition costs possibly bringing more poverty, etc.
Also, it's important to note that the study, according to Monica85, mentions that MTF transsexuals die sooner than cis men. MTF transsexuals doesn't mean "people on HRT". So in theory, the study probably includes many non-transitioned MTFs which, first, reduces the ratio of people in the study who are on HRT by a significant amount, decreasing its importance, and second, factors in people who have gender dysphoria but haven't (medically) done anything about it, which would increase depression's importance as a death factor in that data.
Finally, as I've already said, I believe that female hormones, versus male hormones, reduce the chance of developing quite a few cardio-vascular illnesses, for example. One cause among others would be that a female-type fat distribution pattern generally has less abdominal and visceral fat, resulting in a lower waist measurement, than a male-type pattern. All three of these are strong risk factors for various cardio-vascular issues.
In my very own opinion, if you combine the reduction of these risks, the increase of venous thrombosis/etc. issues, both from HRT, as well as the guidance of quilified professionals who would minimize to the extreme the increase in venous thrombosis/etc. risk, the risk difference between HRT and non-HRT is probably zero.