According to original paper: 
 
fluo-3 was chosen as a more promising candidate for general use;  
 
fluo-2 is much more sensitive to pH changes than fluo-3 that can compromise [Ca2+] measurements; 
 
fluo-2 did not show in that paper better sensitivity to [Ca2+] changes vs. fluo-3; 
 
I expect TefLabs provides some information about how fluo-2 became better than fluo-3 20 years after both probes were sinthesized and compared and fluo-3 was proved to be better. 
 
I really wish fluo-2 works better than fluo-4 but I need more evidences. Thank you!

Dmitriy Thank you for your comments 
 
JPY Kao provided the following in regards to your comments about the 1989 JBC paper: 
 
"There are three important points that were either not known or not considered when the 1989 JBC paper on the fluo and rhod indicators was published: 
 
 
 
1) The importance of pH sensitivity of fluo-2 was somewhat exaggerated in the 1989 paper. Since the perfect or ideal indicator should be completely insensitive to pH, and fluo-2 had a slight pH sensitivity, fluo-2 was judged to be imperfect by Tsien et al. In reality, pH is very tightly regulated in cells--- even when extreme conditions are used, cytosolic pH deviates from the resting pH by at most ~0.2 pH unit. So during a normal experiment, where pH of the extracellular medium is kept constant, the intracellular pH remains essentially constant, so the pH effect, if any, is very minor. 
 
 
 
2) When acetoxymethyl (AM) esters are used for loading indicator into cells, fluo-2 loads a lot better than fluo-3 and fluo-4 (a manuscript containing this result will be submitted shortly). This is partly because the solubility of fluo-2 AM in aqueous medium is higher than the AM esters of the other 2 indicators. More loading means the loaded cells are brighter. This also means that the concentration of fluo-2 AM in the loading medium can be reduced and good fluorescence signals can still be obtained. 
 
 
 
3) On laser-scanning microscopes, fluo-2 is operationally brighter. The optimal excitation wavelength of fluo-2 is closest to the 488 nm laser line used in scanning microscopes. So fluo-2 ends up being better excited by the laser than the other indicators. This is another reason why fluo-2 can give better signals on laser-scanning microscopes."