Ok, the folks at Kanile'a suggested that I should use their original strings. They will send me a new set along with a non-compensated saddle and charge me $28 for it.
While I was waiting for their reaction (for almost a month), I made some more experiments and thought that the results might be of interest here:
Out of the three Kanile'a ukes I have at hand, the Concert has a non-compensated saddle that brings the pivot point on all strings as close towards the fretboard as possible. The Tenor and the Baritone have the same compensated saddle, bringing string number two (E on Tenor/B on Baritone) and four (G/D) as close towards the fretboard as on the non-compensated saddle, while string one (A/E) and three (C/G) are set back by about 1.5mm to the middle of the saddle. In other words, these saddles provide more compensation for the 1st and 3rd string. You can see that on the picture in my original post.
My experiment, however, shows that all three instruments have better intonation with the non-compensated saddle (using Worth Clear on the Concert, Fremont Blacklines on the Tenor and D'Addario T2 on the Baritone). This makes me wonder why Kanile'a compensates their saddles the way they do? For re-entrant tuning (which is the way they string up their Sopranos, Concerts and Tenors), wouldn't you expect string four being close in compensation to string one? For linear tuning, on the other hand, especially when using two wound strings (as they do on their Baritones), I would think that you'd have to compensate the thicker of both the wound and unwound strings more, hence setting back string four and two -- which is the exact opposite of their saddles.
Any idea as to why their saddles are compensated this way? I must say that I'm puzzled.