It is good to take a deep swim in ACES. The whole colour management topic is an important one and at first glance can easily throw you off balance.
Will you indulge me in a brief splurge about colourspace?
It is worth considering what makes up a colourspace. There are basically three things that dictate a colourspace.
The white point. I generally ignore this but I will mention it so that you can ignore it. There are only really two that need to be considered so this bit is easy. It is probably very badly summed up by saying warm or cool. But that is how I like to think of it. You get most colourspoce hitting the D65. Your sRGB, rec.709 and rec2020. Some hit D60 like ACES2065-1. D60 is a bit warmer than D65 but generally I don’t worry about the white point.
The next one is important. Gamma. For now lets consider a log curve and a gamma curve to be the same sort of thing. Or lets think of it the other way. Is it linear or none linear. Log-C. S-log, sRGB all have a log cure (or gamma) that are part of the colourspace. Linear, like ACEScg, does not have a log or gamma curve.
Finally what most people think of is colour gamut. This is often referred to as the primaries. It refers to the amount of colour available to display. Or where the Red, Green and Blue primaries sit of the Spectrum Locus.
Everyone’s favorite, the Spectrum Locus
Jumping around from one colour space to another is made easier by ACES. That is the best take away from my long splurge.
ACES isn’t just a linear colourspace with a wide colour gamut and a white point of D60. It is also a collection of interchanges. It has the profile of cameras, display devices like monitors and projectors. It provides you with a destination to head to from one type of camera so that another type of camera can be used together and a way of transforming out so that you can hit a certain display like a television or a cinema projector and have it look the way you intended.
Phew! I am not on a salary with ACES in case you were wondering.
So that was a long intro. Sorry 'bout that. I’m sure you know all this.
Lets get to your question. Graded rec.709. This has been my world for most of my career and unfortunately the bad news is that once graded, the chain has been broken. Even if you knew what camera it was shot on you have no way of accurately figuring out what the grade has done to the image. Warm or cold. Gamma up or down. It has been worked beyond all hope.
But there is hope.
You might not have the High Dynamic range available or the wide Colour Gamut but you can make use of the linear aspect of ACES. It has excellent tone mapping that will make ugly clipping much harder.
I will not bore you any more with the advantages of compositing in a linear colour space but even with your low dynamic range, graded rec.709, you can take advantage of it.
What you are looking for when you use the colour mgt node is the viewing monitor to look like nothing has changed. You want your F1 input to look exactly the same as the F4 result. If it doesn’t you are either doing the conversion wrong or your viewing LUT is wrong. Simple as that.
If you are going to use ACEScg as your working colour space then the first thing to do is have your viewing LUTs correct.
Next you need to use the ViewTransform. Originally conceived for going from high dynamic range to small dynamic range. Make use of the Invert button and go backwards from rec.709 to ACEScg.
This is a bit of a hack. You haven’t really given your footage any more information. There aren’t any more colours being displayed but you are now in a linear colourspace.
The way this transform makes your footage linear is special because it not only removed the 2.4 gamma curve from the Rec.709 footage, it also did an inverted tone mapping of the footage to extend the highlight and shadow values.
You should find that your linear footage has values that go beyond the range 0 - 1. This has a huge effect on how your linear footage behaves when you composite.
It is still a hack. Highlights will not get any of the clipped information back. In fact those bright colours will be mapped to some very bright values. But you can comp your CG and take advantage of realistic blurs and then invert the ViewTransform getting you back to your graded rec.709 footage.