I've been thinking about this on and off for a while, and Crazyace brought it up at one point over at Sega-16, but I don't think it's come up on Atariage.
Comments from Jaguar designers/engineers John Mathieson and Martin Brennan on the Flare 1 design:
http://www.konixmultisystem.co.uk/index.php?id=interviews&content=john#start
John Mathieson:
We were impressed when the Amiga came out because as a computer it seemed to fit that role perfectly. But, we believed that the Amiga was over engineered - it could animate with sprites and with a blitter, but why have both? We figured that if you could implement a system that could animate the screen then you only needed one set of hardware to do it.
http://www.konixmultisystem.co.uk/index.php?id=interviews&content=martin#start
Martin Brennan:
We based a lot of it on the Amiga as well. That was a great influence - they seemed to have designed a very powerful games machine - if not a games computer then at least a games capable system. They had Commodores purse to design that - and it showed - there seemed to be several different ways you could do the same thing on the Amiga which struck us as overkill. We felt there were lots of instances of doing the same job - but one way - the best way, and it would end up being cheaper. Plus there were certain things we didn't like about the architecture - it was bit planes architecture rather than a bit per pixel, so we felt we had some ideas that would make our lives and the game writers lives that much easier.
When you say was the Flare one derived from the Loki - I think you've got to say that Loki and Flare one were derived from the Amiga, so you've got to pay tribute to those guys.
Brennan's account of joining Atari's Panther project:
Throughout that period we did consultancy work - we did work for Amstrad (we designed a fax machine for them and a hard disk controller). At one point one of the guys from Sinclair joined Atari. He had worked for Perihelion - Richard Miller. He became a director of Atari in Sunnyvale and he had a project called Panther - It wasn't called Panther when I joined. Panther was the name of the car my wife had just bought, a Panther Kallista and the chip had no name and I wanted to give it a handle - so it was called Panther.
The design and specification had already been started, and they said "somebody's left - here's the concept" and it was only the video part of the chip - there was no sound.
It was a novel video architecture that allowed you to create windows of different sizes and different bit depths. Essentially you didn't have a frame store - it was a composite of frame stores - a kind of smart video frame store. It would have allowed a great deal of sprite style animation. Sprites in general in those days would have been of a fixed size e.g. 16x16. The games looked 'spritey' because of that, this would have been quite an interesting departure. I wasn't keen on it, but I designed it and the chip was built.
But while I was over in California in '89, I actually convinced the bosses at Atari that 3D was the way to go, with the experience we'd gained on Flare one - if you didn't just do flat rendering, but shaded rendering you got a 3D appearance.
At the time, I was seeing pictures in magazines where computers were rendering photo realistic 3D wire meshes and I said "these are static images, but they only contain a very few number of polygons - we could take that, animate it and you could produce a game that was a quantum leap away from the current games".So the Jaguar project was born from the Panther project.
In essence Atari looked at the Panther and looked at what we were promising for the Atari project and said can the Panther project.
So with all that in mind, it seems rather odd that they ended up using the Object Processor in the Jaguar rather than a more streamiled blitter-specific design. WIth both the Object Processor and Blitter, you've got a similar situation as to what Brennan and Mathieson both criticized the Amiga for, with effectively redundant hardware adding unnecessarily to cost (or, conversely, taking up valuable chip space that could be used for other things).
Furthermore, Brennan expressly states he disliked the general Object Processor concept when working on the Panther, which makes the inclusion even stranger.
I could see the logic behind using some of the design elements of the Panther, like the variable color depth textures/objects with color expansion done via 8-bit offset (which the Jaguar OPL does as well) and using such features in the Jaguar Blitter, but not retaining the object processor architecture itself. (which takes up far more chip space than a minimalistic framebuffer controller would)
I could also see including the OPL had the Panther actually been released (aiming at compatibility while making efficient use of existing hardware), but that's not the case either. (the Panther was cancelled before Jaguar development really got underway)
A more powerful and flexible blitter would have been more useful (and easier to program/port games to) for a good number of different types of 2D game styles and effects, and also more useful for 3D. (having added logic/buffering for fast blitter scaling/rotation effects would be nice for 2D, but also mean much faster texture mapping for 3D . . . and much faster pseudo 3D texture effects like "mode 7" style skewed planes)
Hell, with the object processor gone, they might have even had enough chip space left to add some of the features that the Jaguar 2's blitter got, like a texture cache (enen a small one), double buffered blitter registers or maybe even distorted triangle/trapizoid rasterization.
That, and potential efforts that could have been put into the GPU as well. (aside from bug fixes, there's things like including a cache controller and allow the scratchpad to become a cache -even a relatively simple/primitive direct mapped cache scheme should have been pretty useful . . . after all the R3000A in the PSX used that 
)
Or conversely, they could have used the savings to just focus on lower cost and overall system simplicity . . . and potentially more quickly working out the bugs. (though the blitter should at least have had a few added features to make it more useful in leu of the OPL, like support for indexed color textures)
Actually they could have even cut out the big chunk of TOM used for CRAM and just used direct color textures along with 4-bit (15/16 color) color, or allow 1/2/4-bit and just 16 CRAM entries for blitter color expansion, and do away with 8bpp textures and 8bpp framebuffer modes. (unless they REALLY wanted to do 8bpp stuff . . . which could be useful given the potential bandwidth advantages for an 8bpp framebuffer)