The current should drive an optocoupler (effectively a LED), so in my opinion the 0.5mA is not going to make a big difference to how the LED lights up in the coupler.
Have you checked that the opposite (ie, sending 0xff continuous) gives a current that is at least close to 0mA? That couls also be a problem as the LED would stay at lkeast partially lit.
One other thing that may be worth doing is cheking the specs on the Roland cable to see what is expected there. You may have done this already but there is not info on these posts.