After several weeks of trying to solder the RF amplifier chips I eventually concluded that the techniques that I was using to accurately position the chips was not up to scratch. I tried microscopes, magnifying glasses, magnifed eyeglasses that you jam into your eye socket - all to no avail.
Using the square drawn on the silkscreen was hopeless as the board manufacturer could, or would, not print the silkscreen accurately and this route proved to be a dead end.
I did elongate the PCB pads slightly so that they were visible, but the pads on the chips themselves are not visible at all so there was no possibility of lining them up at all - another dead end.
After a while, just by chance, I was able to get the two chips, the LNA and VGA working together by connecting up different boards and when they worked, they worked really well. This was nice as it effectively ruled out any incompatibility issues. Unfortunately, the laws of probability did not allow me to get the two chips working on the same board ..... Not yet anyway.
So, I needed a fool proof way of aligning the chips and eventually came up with the idea of using some top layer pads as markers instead of being used for electrical connections. In the diagram below there are four tiny red squares, one of which has a blue arrow pointing towards it. These pads are placed exactly corner to corner with the main chip so that the chip can be nudged into position by focusing a microscope on each corner at a time. This should enable positioning to about 0.05 mm which is absolutely fine for chips with a 0.5 mm pitch. Now just need to get another batch of PCBs made :(
It's worth noting that trying to use a microscope to look at all the corners of the chip at the same time will inevitably be inaccurate due to the marker squares being partially obscured by the chip, particularly severe if the chip is quite tall.
The first diagram below shows the chip in place, as a solid white square: