Don't know for sure, but I have experienced the same type thing on tig torches using gas lenses. Lots of aluminum buildup on gas lens screen.
I have found, that when starting, if enough current is not used on starting, the arc tends to be very erratic. This erratic arc can volitize some aluminum, with it's low melting point.
Keep in mind, too, that an erratic arc can be traced down to worn consumables (collet and/ or collet body).
As an experiment, you might try a new collet / collet body. In addition to this, make sure you are using enough current to "light off" the tungsten. Nail it (current wise) when starting and back off to the current level required. If you can't make a puddle in aluminum in 4 seconds or less, you are too cold.
Be sure, too that you are not using too large a tungsten. A too large tungsten will amplify this problem. Low current/tungsten diameter ratios can always cause hard starting and the problems you mention. Also, a cold tungsten does not start as well as a hot one does. A quick heat-up on a clean piece of copper or stainless might help.
There are a number of size/amperage charts available on the net from various mfg. If you come up empty handed, email directly and I can forward a chart.
Anyway-keep us posted. Hope this helps a little.
brande