It applies to any wideband amp in the signal chain prior to the channel tuning - if the 4G signal causes it to saturate, all other signals will be wiped out.
I am aware that the aerial itself applies a degree of band-pass filtering, but the cut-off is not very sharp so there will be some kind of pick-up even from a Group A aerial (the A band goes up to 600MHz, 800Mz is less than half an octave above that) depending on the direction of the 4G signal relative to the TV signal and the polar (ie directional) response of the aerial at the frequency in question, and any break-through in the cable etc. It must be understood just how strong a mobile phone signal is: compare the size of aerial used to receive TV with the size of aerial on a mobile device, and of course the aerial on a mobile cannot be directional which means it needs a stronger signal still.
The problem with amplification is two-fold: the easiest to understand is clipping. The amplifier has a certain maximum output voltage excursion, above or below the average output ("zero"). If the input signal is so great that the output signal would exceed the maximum excursion, the signal peaks get clipped at the maximum. However, if the signal that is being clipped is the unwanted 800MHz signal and the signal you actually want is a superimposed ripple on that, the clipping wipes out the wanted signal.
Problems occur before the amplifier reaches clipping though. Amplification theory relies on the amp being "linear" - that is, the output is exactly proportional to the input. Unfortunately the linear region is much smaller than the full output range of the amplifier (limited by clipping). I won't go into the maths, but as soon as any non-linearity is introduced into the amplifier input-output curve, signals start interfering with each other instead of passing through essentially separately. If you have signals going in at (say) 600MHz and 800MHz, 600MHz and 800MHz come out but also 200MHz and 1400MHz (the difference and the sum).
This is called intermodulation distortion (and is how the heterodyne stage in a receiver works to reduce the "tuned" frequency to a much lower fixed frequency into the tuning and detector stages). The problem is that there are not just two signal frequencies going in, there's a plethora, and some of the intermodulation products will be at the same frequency as your desired signal. A strong local signal, even if out-of-band, could over-drive the amplifier so that it goes non-linear and then the intermodulation produces a mush to try to pick the signal out of (ie makes the signal-to-noise ratio worse).
This shows what the problems are. To what extent they will affect anyone in particular depends very much on the local conditions. The solution is to eliminate the 4G signal before it passes through any amplifier, including the mast-head. Eliminating it after the mast-head amp or other distribution might work if the problem is occurring at the receiver, if so - fine. If a 4G operator causes anyone's TV reception to fail, they are responsible for sorting it out so don't be afraid to complain - first to the operator and then to the regulatory authority if necessary. Solutions might include simple filtering, replacing the aerial with a more selective one, pointing it at a different transmitter, or in the worst cases offering a cable or satellite alternative to Freeview.
Having thought about it, provided the 4G signal is sufficiently off-axis, my preferred solution is to link two aerials in a phased array positioned precisely to reinforce the Freeview signal and cancel out the 4G frequency/direction, with no mast-head amp, then kill any residual signal with a notch filter at the bottom of the downlink prior to distribution. This isn't going to work if there are two 4G signal sources locally!
