I have recently been working to repair a Sansui 551 that also has the F-2591 board rather than the F-1509 as documented in the available service manual, and I think I've gotten to the bottom of why there are issues setting bias on this version of the board.
In the F-2591 revision, for some reason R87 is fed from F01 (the left channel power amplifier fuse) and R86 is fed from F02 (the right channel power amplifier fuse) rather than being fed directly from the unregulated 52V input (which itself is already fused by F04 and F05). I'm not sure why this change was made (maybe circuit board layout reasons?) but the end result is that there's a static load of 20-30mA (I expect this is somewhat dependent on the exact unregulated voltage) on the left channel fuse in particular, which completely throws off the documented bias adjustment procedure.
My problem turned out to be something else (corroded capacitor terminal on C38 resulting in basically no amplification in the right channel power amplifier stage), but I got led astray by this strange bias difference while troubleshooting. I tried moving R87 and R86's feed points to the unfused side of the unregulated supply, and sure enough I was able to adjust bias current as expected for the left channel now in addition to the right channel.
Rather than modifying the circuit though (there might be some other reason they redesigned this part of the circuit), the better solution for anybody encountering this problem is to use an alternative procedure for adjusting bias. My experience is that the amp should be set to aux input with no signal (rather than a 1kHz sine as documented in the guide) for the procedure since depending on your meter an incoming signal can affect the current read-out with modern fast digital meters. Volume control seems unimportant; with no input signal, the bias value is basically unchanged from minimum to maximum volume.
Once the radio is on and warmed up, set to aux input with no signal being fed in and speakers unhooked, measure the voltage drop across both power resistors for each channel (this is R71/R73 for the left channel and R72/R74 for the right channel). To clarify, clip your meter leads to the outer ends of each resistor pair, rather than clipping both leads to two terminals of the same component (this means you're measuring voltage across a larger resistance drop from both resistors in series, giving more accurate results).
The nominal bias current value of 20mA from the manual works out to 18.8mV across these two resistors in combination; adjust the bias potentiometer to get a value of about 19mV and then do the same for the other channel. Throughout the procedure, the fuses should be left in their normal position and don't need to be touched.
While it may "work" to set the bias to a higher value as described by MWH901, I expect this is actually creating a bias imbalance between the two channels (since only the left channel has substantial extra load current on the fuse in the revised design) that might affect sound character or distortion, and more importantly it's putting the power regulation circuitry under additional load and those components are already the hottest-running ones in the whole amp by far. For anybody who comes across this, setting the bias by using the voltage drop across the big power resistors in the output stage is definitely the safer option, and the procedure should (I think) work equally well regardless of what revision of the board your Sansui 551 has.