Why does not the conclusion that the fish perceive blue as the main color of UV stimulus follow from "MAIN SENSORS OF THE UV LIGHT WILL NOW BE THE BLUE CONES"?
Because of "wavelength discrimination.
The image below might help:
In the image in the top left, when a photon of 500nm is absorbed, it can be absorbed and expressed by many different cone opsins. Draw a line from the 500 straight to the top. Every opsin it passes through will be triggered to a greater or lesser degree. However, because 500nm is visible green, the intensity of each opsin should still register on the visual centers as green, with different intensities.
As well, note the peaks of each cone opsin type. While each cone opsin type has a unique peak, they all have a secondary, lower peak in the UV. So, when any of the cones receives a UV photon of 350nm (the left axis), the visual center should register it as UV. This is wavelength discrimination, the essential element of color vision. And Flamarique's statement holds true. If you look at the graph, the opsin expressed by the SWS2B (blue cone) reaches 37.5% at 350nm; much greater than all the other cones but SWS1 (UV specific).
Does the previous make any sense? A good paper on color vision is
I won't try to tackle the difference between sight and perception; however, there is a chapter on the subject in a certain book which shall remain unnamed.
Flamarique and Cheng stand pretty much alone in thinking that the UV-specific cones in the dorsal-temporal region revert to blue-specific. They have both made many valid contributions to trout vision science, but they have been unable to prove their conclusion and the other camp provides better, IMO, arguments.