Welcome to the forum,
Post some of your parameters (ie. size of your CH machine/amps/volts/wirefeed speed/gas flow in cfh/type of gas(es)/ joint config(groove,fillets?)position orientation(flat, horz, vert?)/ type of base materials(carbon steel A36?) and thicknesses/ wire designation (ER70S-6?) and size(.030" , .035", .045"?) /etc...etc...)
Sounds like one or more of these are out of whack.
A few things to think about...GMAW is a little fussier about cleanliness than FCAW is. You'll notice that mill scale plays a big part in this. FCAW will dig (penetration into base material) and clean better than GMAW, it will float a lot out in the slag. You'll find your GMAW filler just laying on top looking like a pretty weld but without any penetration if you are not watching what you are doing. If you are getting a lot of silicon on top of your GMAW beads, likely you are not cleaning enough mill scale off. It should produce a bead with almost nothing to chip off, unlike FCAW which has a thick layer of slag.
I suspect your wirespeed(amp)/volt relationship is your problem with the wire balling up. The "flame" you see is most likely the plasma stream carring your melted filler across the arc into the puddle, it kinda looks like a inverted icecream cone shape. Once you get your puddle started you need to keep your wire feeding out on the leading third of the puddle. If the puddle is running out ahead of you, then you are likely to have problems with penetration.
post back with more of your info,
John Wright