Tosfot is VERY bothered by the fact that this Ika d'Amrei is exactly the opposite of the previous version. They spend almost 12 lines on showing how this just can't be, that it is totally unprecedented, and offering an alternative which surprisingly the gemara didn't do, etc. (And then proceed to give their resolution).
However, a little later, on 26b, the gemara presents another Ika d'Amrei concerning Muktzeh l'Chatzi Shabbos and it too is exactly the opposite of the previous one - and there's no comment from Tosfot at all! (there aren't even any Tosfot on the page!) 1) Here IS a precedented case, in the same masekhet, of an 'opposite Ika d'Amrei', and 2) How come Tosfot isn't even bothered just a little bit here?!
That is a very nice point, and it is good that you noticed it!
Why is Tosfos making a point of the fact that 'nowhere else in Shas' are variants of a similar nature recorded. Who says variant versions of disputes in different places in Shas follow similar models? What is bothering Tosfos is not the uniqueness of these variant versions, but the logical inconsistency implicit in them.
The first version of the Gemara demonstrates that there is more reason to require separating Ma'asros from bundles of Kitniyos than there is from Shibolim. Yet in the second version, the Gemara seems to consider the opposite approach more logical - there is less basis to require separation Ma'asros from bundles of Kitniyos than there is from Shibolim. Which type of bundle is more 'susceptible' to Ma'asros - Kitniyos or Shibolim? Logic must dictate one way or the other, so how can there be variant versions of such a discussion?
This inherent logical inconsistency of the Gemara is what is bothering Rav Shimshon of Kutzi in Tosfos. Certainly, we can find many instances in Shas where two 'Ika d'Amrei's argue whether something is Mutar or Asur, Chayav or Patur. But we cannot have to 'Ika d'Amrei's arguing over whether A is more logical than B, or B is more logical than A.
Kollel Iyun Hadaf