I don't see nay mixer at the input, only 2 LNAs. It would make sense to be a hybrid coupler as they have a hybrid also at the exit of the LNAs. This would suggest the IP3 was too low and they needed to back off the power? It still doesn't make sense why to use this method at the receiver, again you loose on the noise figure. It might be possible that this is a universal board and each of the units cover one band as they are older parts?
After taking a better look I noticed all three ICs are the same type of LNA/VGA. This means they are all wide-band and there must be a reason why they decided to use that configuration at the input and it isn't the IP3 as this would affect the second stage even more.
I checked the RX IF, it seems to be around 10GHz, so my frequency plan goes out the window. It seems like the TX is directly upconverted and the RX has a heterodyne architecture with the IF being at 10 GHz and all the VGAs are there. The mmW parts are all fixed gain LNA ICs. This also makes sense.