If a model does not reflect reality, the model is not accurate enough for that situation. Consider some of the minutia in the workings of light.
First an alternate model for the fiber: As light passes through, some of it passes straight through, some is "absorbed." (quotes because I suggest including reflection and refraction in here as well as what is truly absorbed and transmitted at a different wavelength). Some of the "absorbed" just changes direction (reflection, refraction), some is truly absorbed and changes both direction and wavelength (some likely infrared), and some becomes heat. Heat increases temperature of the material and then transfers to materials of lesser temperature through conduction, convection, radiation. I suspect you have the heat transfer portion's model down, but the optical model needs to be refined based on the wavelength of the light and the atomic structure of the glass. Note that atomic structure at these levels would also be a function of temperature, and that light can interact with light making the wavelength fairly complex too. You can get as detailed as you need to and incorporate measured properties as necessary until your model matches reality. In fact, the very measurement that revealed the inaccuracy of the model can itself be used as a model (albeit not a very useful one with just a single point) for what would happen under those exact conditions.
Second, an alternative model for the sheath. (Even if there is no sheath, the outer surface exposed to air will have a different structure and different optical properties.) You can model it similar to the fiber, but with different properties. For example, the usual sheath will let significantly less light pass through compared to the fiber so that most of the light reaching the sheath will fall into the "absorbed" category. In other words, probably generating more watts of heat per watt of light compared to the fiber.
Finally, you may need to model the distribution of light entering the fiber, but this is unlikely unless you are looking at sub-mm of fiber or some fancy crystal structure.
For simplicity, some light goes in (30W), some light passes straight through (96.666%). The rest eventually makes it to the sheath, although a lot of it is still light that turns to heat at the sheath!