Like I said, a lot depends on what you consider to be acceptable. I think about +/- 0.7dB levelling is typically possible (at the end of a short RF cable) with a fairly modern mid range 1GHz sig gen. So if this is acceptable then this should be fine. If you have a decent RF attenuator you can improve the source match of the sig gen but you would ideally have to know the attenuator was healthy in terms of VSWR and flatness to 1GHz.
For a bit of fun, I powered up an old Marconi 2019A sig gen. This operates up to just over 1GHz. This sig gen cost me about £120 over 25 years ago and I rarely use it. However, it does support external AM modulation with DC coupling.
I also powered up an old Anritsu power meter I bought in the early days of ebay. This cost £200 complete with a thermocouple sensor.
I also used my homebrew levelling amp to level the output of the 2019A sig gen via the external AM mod input. The levelling amp is just a cheap opamp in a box with a potentiometer that sets the output level. This was made from scrap parts and effectively cost me nothing.
The sensor cal shows about +/- 0.05dB for the correction factor all the way to 16GHz.
I also used my homemade version of the 11667 splitter. The resistors cost less than £0.50 to make this. It also needs three RF connectors but these can be purchased for about £5 each.
I then set the output level target to -12.00dBm using the levelling amp. I could have set it to anywhere from about +10dBm down to about -20dBm.
The levelling amp was able to level from a few MHz through to the 1GHz limit of the sig gen. The power meter displayed -12.00dBm at the output of the splitter on every test frequency. But then it should do this because it is in an ALC loop. In reality, the accuracy will be limited by the quality of the splitter and the accuracy of the thermocouple power meter even though the display shows -12.00dBm at every frequency.
The total budget for the above is less than £400. If I had used a 4GHz sig gen then it could have been used to measure 4GHz scopes with low overall uncertainty.
The risk with doing this with 'just' a sig gen is that unless you have a way of verifying that the output is within specification you have to hope that the sig gen levelling is OK.
The Marconi 2019 has fairly good levelling up to 1GHz but I'm unsure of the source VSWR. An external attenuator can improve this, but again, you ideally have to have something to try and prove the level is flat. If this means buying a power meter then you might as well use the same levelling system as me.
A decent homebrew diode detector could just about work to 1GHz if you wanted to (quick and cheaply) prove the sig gen was flat to 1GHz +/- 1dB but this would require a sig gen with low harmonics.