The Ultimate Black Mage Resources with Damage Calculator

[Sleigh]

I'm not seeing how you're deriving "10-20" DPS from using Thunder III over II.
As I understand from what you're saying;

A) Fire > Blizzard III > Thunder III > Fire III is giving you 10-20 more DPS than
B) Fire > Blizzard III > Thunder II > (wait .5s) > Fire III.

Which is frankly nonsensical, but I do understand where because of the mana ticks that the additional cast time on Thunder III is irrelevant, I'm just pointing out the numbers will be no where near as high as you make them out to be.

Getting an early mana tick however and using a Thunder I instead of II to capitalize on that still only gets you back in to Astral Fire less than half a second quicker, which could potentially give you DPS gains over a long fight, and most likely where your "10-20" DPS increase is being noticed.

Something that does strike me as an interesting proposal would be when we approach extreme amounts of spell speed, i.e. BiS from this spreadsheet & spell speed food; if it wouldn't just be better to start always using Thunder I, and then filling an any addition time with a Blizzard or Scathe depending on procs. Since this allows you to always spend a minimal amount of time in Umbral Ice, without delaying your Astral Fire phase. & Thunder I + Blizzard is going to be a net DPS gain compared to a Thunder III.

I don't necessarily get 10-20 more DPS from Thunder III alone, I get it from making sure I use the right UI3 filler over trying the old way of Thunder II every UI3 phase as worked so well in 2.0 but can lead to upwards of 1s of down-time in 2.1 - I've seen Thunder III fail to give a full two MP ticks before activating AH3 so Thunder II can definitely leave a bit too much empty space after it

That being said, if you don't want to take my word that it's an increase in DPS, feel free to test something like a 5 minute test using Buttons vs HQ Mulled Tea, both at 241 base Piety, plus a SCH in your party. If the Button set wins I'd be more than willing to do some more testing myself.