An RTD installed on the end winding of the machine is going to measure a dramatically different temperature from one installed between coil legs in the stator slot. As a rough estimate - a difference of roughly 60-70 C for a normal Class F rise design. One of the reasons for this is that the detector in the end head of the coil gets exposed to the surrounding air temperature as well (which lowers the average reading significantly) and there is even more convection effect by the amount of air moving past the detector.
An RTD element insulation suitable for your machine voltage level of 13.8 kV is going to be a thickness of 4.5 to 6 mm (0.188 to 0.250 inch). This could - in theory - be mounted on the gradient tape portion but it is not advised to do so. Likewise it is not advised to mount it on the exposed length of semi-conductive ground wall tape either. The down side is that the device will likely have some length to it - which means it isn't really a good fit for the exposed coil nose area either.
Best bet is to go with a "button" type detector, rather than a "stick" type, to allow mounting on a surface with shorter length. If you can, tuck it out of the main airstream and minimize the material - including any fixative substance like glue or varnish - between the device and the coil.
For long-term reliability reasons, don't put it on a line coil if you can help it, or on either of adjacent phase lead coils. This potentially exposes the device to multiples of line voltage, which might be enough to develop a voltage in the detector which could then ruin your data acquisition unit. It is recommended that you install at least one per phase (and for a larger machine, probably two or three per phase) - and that you reconfigure your trip/alarm setpoints to correspond to the new measured data. Because of the temperature differential measured between the OEM "slot" detectors and your "surface" detectors, you're going to have to use one set or the other - not both.
The purpose of the teflon (or similar material) shielded jacket is to provide additional mechanical stiffness to the detector wiring (it is pretty flimsy when you get right down to it!), and to prevent inadvertent circulating currents/voltage from transferring from adjacent "power" conductors to your "instrument" wiring.
Converting energy to motion for more than half a century