Two HF/VHF/70cm antennas coming within 1m of each other is very tricky, especially in your situation. I would maximize the separation as the first measure. Your endfed is horizontal, so that much is good, but when you have to come down by 47dB to be safe, that alone is not sufficient.
First, measurement. NanoVNA is the wrong tool for this. You just need a very sensitive wideband micro power meter that can measure down to -20dBm or so. You could build one using a chip like AD8310, AD8319, or maybe AD8314, coupled with an attenuator. (My knowledge here is a few years old - check the latest spec sheets when selecting your chip and the design.) If you would rather buy something more versatile, TinySA Ultra would be my first choice, also with an attenuator. Start measurement with the lower power transmit and increase the power until you can get reliable readings. Then calculate the "link loss" between your antennas.
The high-frequency behavior of an end-fed half-wave antenna is very unreliable/variable and susceptible to many things. I generally do not recommend using a half-wave antenna on higher bands if performance is any concern. In this context, that gets back to you as an uncertainty or unpredictability.
One good thing is that the 49:1 type transformer is very lousy, especially at higher frequencies, so you may already see attenuation when you transmit at 50MHz and even more so at higher frequencies. But that is also unreliable and unpredictable because those transformers are never designed for higher frequencies, and self-resonance and other factors can come into play.
Therefore, my recommendation is to insert an LPF with a cutoff frequency of 33 MHz or so in front of your SDR receiver. This is fairly easy to design and construct. Additionally, I would add a warning system using AD8310 or 8319 to cut off the antenna connection via a PIN diode or a small signal relay(1) when the receive antenna power exceeds 0 dBm. This should also be fairly straightforward to design and construct. (I would just do strictly analog electronics design rather than a microprocessor, so that the cutoff can be fast and reliable.)
(1) In fact, for this application, you could just use a regular switching diode since the signal amplitude is tiny.
ADDENDUM
The LPF can be as simple as this. It's a prototype I built (just used standard capacitors and inductors as wound, without any tweaking) and measured. This would be good enough for this application, as it has 50dB attenuation at 52MHz and 145MHz (though that could be further fine-tuned, as the filter is based on an elliptical function). (S11 is not ideal but good enough for a lot of applications, including this one.)
(This is something I did for a completely unrelated project in 2023, and I'm showing it only as an example... it is probably good enough for your application as-is, but it has room for customization/improvement from an overall system performance perspective.)
