Homodyne vs Heterodyne for DIY VNA

Hello all, previously I finished a project in which I basically made a simple VNA designed for analyzing pulsed RF amplifiers. I used a AD9959 quad channel DDS to generate the RF and quadrature LOs directly (my frequency range is 10-150MHz). I used homodyne conversion, so therefore the RF and LOs were at the same frequency. For this project, I was only interested in measuring forward transmission coefficients (S21), and didn't really care about reverse transmission (S12) or reflections (S11/S22). This figure shows my basic homodyne architecture:

However I wanted to take things up another step and use a heterodyne downconversion to eliminate offset errors due to baseband detection. Also I thought it might reduce the required hardware by eliminating some mixers and ADC channels. I imagined it like this second figure:

However I think that it"s not so clear when I think of how the phase measurement works. One advantage with the homodyne design is that since the RF and LOs are at the same frequency and are always phase locked (due to being generated by one DDS), I could measure the phase of S21 just by looking at the I/Q ports at the receiver (port 2). However, with heterodyne downconversion, the LO and RF are at different frequencies so obviously there is no phase lock. In this situation, is it correct to say that in order to get an absolute measurement of phase, I have to also sample the power out of port 1, and sample it after downconverting it using the same LO as used on port 2? This means I still need two mixers, two ADC channels, a splitter for the LO, and a DICO on port 1, like shown below:

So after all is said in done, is there actually going to be a significant advantage to moving from baseband to heterodyne? I"ll get rid of baseband offset errors, but then won't things like matching between IF filters become a problem? And aperture delay variation on the ADC channels?

the last diagram is the one that the traditional network analyzers use. you are looking for a phase difference between ADC 1 and ADC 2. so the phasing is more obvious.
the DC offset in the first block diagram IS a real PIA to handle. and the dc offset is also affected by the RF input power to port 2, so calibrating it out is not so easy. you would use it only for consumer type applications, where you do not care about absolute accuracy.

That's understandable but I was hoping to reduce the number of components in the design. So far I've been using minicircuits splitters and passive mixers, and those take quite a bit of space even though they work well, and then I have to worry about laying lots of transmission lines and calibrate for phase skew on the LO paths and such. I'm wondering if there aren't any chips that have many of these elements built in? Like with a PGA > mixer > IF filter > ADC in one package, or multiple chains per package. I've seen some, but none seems to work with RF frequencies below a few hundred MHz. I was hoping to operate down to 20MHz.

Also is there any reason to avoid active mixers in an application like this? Many require much lower LO power to use, which would be very helpful. I'm looking at parts like the AD8342, AD8343, ADL530, etc. And could I use a good clock distribution IC instead of a wilkinson splitter for distributing the LO? I'm just trying to optimize things a bit.