Minimizing Effects of Nearby Metal on 13MHz RFID Loop Antenna Design

I'm getting closer to a useable prototype for my rfid project (usable means can be installed in varied ways, some of which may involve flush mounting in a metal part).  What I've found so far is the best way to design my matching network is to check the impedance with a nanovna at the amplifier, and tune the network to the antenna as it is in the device.  I've built a test enclosure that represents the worst possible configuration an end user could subject the device to, basically a recess in an aluminum block that sits the face of the device flush with the face of the block.  I've noticed a huge difference in the impedance at the amp when the prototype is in this enclosure, but I was able to retune the network to match the antenna impedance when installed like this.  my question is this: is there a way to minimize the effects of nearby metal on a 13mhz loop antenna?  right now I'm running a flat spiral with 6 turns with an OD of 1".  would it be better if it were fewer turns?  more?  etc.  also what's strange is that the measured amplitude of the signal at a known distance barely changes when the device is in the enclosure or out, that seems strange to me because the impedance at the amp goes from 70 to 10 ohm...

Sean,
I'm not surprised by your results. Anytime you have metal in the near field of an antenna, it affects the pattern and the tuning. In this case, the aluminum enclosure detunes the antenna but at the same time acts as a reflector.
This effect won't change much by changing the number of turns. Your efficiency might improve with more turns since your antenna is small compared to the wavelength.

thanks for the input, I'm looking at redesigning the antenna as a cylindrical coil on the inside diameter of a wide torus ferrite, my guess is that the permeability of the ferrite might isolate the field from metal on the other side of it, but still allow for the field to "stretch out" in front of the antenna.  other options I'm considering are an rf amp inline with the rfid chip and matching network, I was able to detune the network to give mediocre read range both inside the metal housing and outside, without retuning, but I'd rather have satisfactory range in both cases.  battery life probably won't be affected too much since the reader is only active a very small part of the time until a card is detected and accepted or rejected.

Increasing the number of turns should improve the range. Another idea is to add metal (either a plate or slightly larger coil) to the back of the antenna to act as a reflector so more of the energy goes in the direction you want. This will also reduce the effect of adding the enclosure.