I have taken apart the FFL505A (3500 K, rosy tint).
A circular die is actually used in the FFL505A, as in the well-known “Yinding” versions. A kind of “bulge” at the edges of the half-dismantled FFL505A chip is noticeable, but this is just an optical effect caused by residues of silicone and phosphor. The chip itself is perfectly square.
Like on FFL350RD there is a glass pane above the phosphor layer. It seems that it was glued to the die with this phosphor layer. The white silicon molding is very stiff and not really flexible at all. It breaks into pieces if removed.
The LED chip of the FFL505A is identical to that of the Lumenpioneer W5050SQ3 (3000 K):
No light is wasted through shading; the FFL505A is simply inefficient due to the phosphor mixture (and maybe LED chip radiant flux binning). I suspect that the special phosphor mixture has some sort of “masking” effect, albeit to a lesser extent than with the white silicone of the FFL350RD, and also falls into saturation much more quickly - maybe due to the red phosphor (KSF:Mn4+ ?) mixed in.