In all the comparative flight test data I saw (and I was at Northrop at the time), the YF-23 consistently outperformed the YF-22, could carry as much, and the production proposal had a more flexible, in terms of carriage, internal weapons bay set up with the builkhead between the two bays capable of shifting fore and aft like the bulkhead between weapons bays 1 & 2 on the B-1B. Where the F-23 proposal fell down, majorly, was on the management side; the USAF not having tremendous faith at that point in Northrop's ability to manage the program. This is the information that was conveyed to the engineers after the post-award de-brief by the USAF.
Oh, the extensive sawtooth shown isn't necessary (take a look at the F-35's inlet, or that flown on the JIST testbed to prove out the theory behind it. The YF-23 didn't need vectoring nozzles, it already was as maneuverable as the YF-22 and the nozzles just add complexity and weight.
Matej, if you want to add an alternate reconnaisance capability that was palletized, consider a shape similar to the lower front of TSSAM, both front and back, faired into the surface contours, and using either one or both weapons bays. Such a shape could carry both cameras (with appropriate LO windows), passive ELINT sensors, and/or low probability of intercept active recce radar.