Hi Maurizio,
Normally if you don't have any problems, you should see a clean carrier in the middle of the screen, without those spurs. The spurs indicate that you have some unwanted modulation on the calibration signal, but this appears not at the generation of the cal signal (not at the BNC output). Most probably somewhere on the analyzer input path (1st IF, 2nd IF and so on). It is a typical power supply/filtering problem, and it is very common for the Advantest analyzers of that age. I also had a similar problem, which leads to user calibration errors. Please see the images below: 001,002,003 is the cal signal in different span and dB/division view (measured with a good working analyzer). In zero span, the analyzer performs like an oscilloscope, you will see the demodulated signal. You can see, in zero span it is a straight line, there is no noise over it. The zero span is very handy, to determine the cause of the spurs/noise. In the pictures B004 through B007 there are the captures of the same calibration signal taken with my defective analyzer (Advantest R3273). In zero span, with 10 dB/division it also can be seen (this should be a straight horizontal line) that there is a ripple. In the last picture I triggered to the video signal (the resolution is 0.5dB/division), so it looks like an oscilloscope image, and is clear that there is an additional modulation detected, the period can even be measured with the help of the markers (the power supply is a switching mode one, so we do not chase 50 Hz or harmonics, the frequency is in the range of multiple of 10kHz). In the end the solution was to replace a lot of capacitors, first all inside the power supply, and also some others in the different modules.
If you have the possibility to use another analyzer, and measure the signal on the receiver path of your R3463, you can trace down where the most part of this noise appears on the path. You have to keep the analyzer with zero span (tuned to 20/30 MHz, depending on the analyzer, and the CAL signal connected to its input), and you can measure along the path (1st IF, 2nd IF, etc.) with the other analyzer the signal (also in zero span). You will find where this additional modulation appears.
This unwanted modulation causes the step amplifier/attenuator calibration failure. During the user calibration, when the IF is calibrated, the analyzer keeps the input RF attenuator at fixed value, and switches some IF step amplifiers and attenuators with small increments (0.1/0.2/0.4/0.8 dB). As it can be seen in images the ripple causes a change in amplitude of the demodulated signal more than 1 dB, which confuses the analyzers measurement/calibration routine. It measures a reference level first, makes a small attenuator step change, and when measures again, due to the ripple, it will detect an over range value. So it will report an error.