Further research into this shows the problem is with division. For example, this equation
$T/1.0
works okay, every pixel is the same as before,
but this equation
$T / $T
changes every pixel to 65535 (unsigned 16 bit integer formatting)
it should have been 1 for each pixel (e.g., 100 / 100 = 1)
There's no problem there. PixInsight natively works with floating point numbers in the range 0.0-1.0
But if you choose to display the pixel values as 16 bit integers then they will be displayed in the range 0-65535
You could equally choose to display them as 8 bit integers in the range 0-255
The answer to your original question:
y = 0.99 *x + 140/x + 318.0
depends on the range of values that your version of x is allowed to take.
I'm guessing your x is in the range 0-65535 and your constants are compatible with the same range, then the PixelMath formula you want is:
(0.99*($T*65535)+140/($T*65535) + 318)/65535
This is because PixelMath values of x are natively in the range 0.0-1.0 and the result of the expression needs to be in the same range.
Mark