Ok, finished testing. Demonstrating Foldypaws's point that it is not light-light;
Code:
Light Elemental (Tu'Lia): Light
293 228 Shell IV
376 376
307 307
271 211 Shell IV
351 274 Shell IV
360 360
378 295 Shell IV
368 287 Shell IV
382 298 Shell IV
364 364
So 10/10 without any resistance at all.
Code:
Helter-Skelter(Abyssea-Vunkerl): Light
1696 848 50%
1696 1696 100%
1988 1988 100%
1901 1901 100%
1670 1837 110%
1863 1863 100%
1132 1132 100%
1502 1502 100%
1223 1223 100%
1602 1602 100%
Interestingly, it was lightsday for all of these tests, hence the 110%... I'm not sure what to make of that.
Since Djinns take a 50% damage cut from all sources except fire which they are also weak to, this indicates that on this particular test, fire was used for resistance purposes, and that which element used depends solely on the resistances of the target.
Now, for damage type when there is no particular preference (Puks are either only slightly resistant to wind, or not resistant at all despite absorbing that damage type)
Code:
Sea Puk (MMJ Staging Point): Fragmentation
508 304 59%
477 257 53%
492 132 26%
339 203 59%
496 297 59%
543 325 59%
313 187 59%
285 171 60%
336 201 59%
529 317 59%
All attacks inflicted damage, rather than healing them as can be seen when creating a detonation skillchain... Further testing on Turul with fragmentation skillchains yielded similar results, with him taking damage rather than absorbing it. This may point towards damage type for L2 and above being 'all-purpose' or 'non-elemental' magical damage, with the element of least resistance only being used for resistance check rather than actual damage type.