Exactly to be expected and WHY you don't use such a rendering intent unless your goal is to simulate the paper white of another process. Absolute and Relative Colorimetric share the same profile table and expect for this handling of white, do not differ whatsoever. So you probably don't want to be using Absolute. FWIW, you should also see this paper and white effect soft proofing.
They share the same table but alter much more than the handling of white. They try to replicate, colorimetrically, the image's colors without change. Rel. Col. shifts everything, not just the paper white, to align L*a*b* 100,0,0 with paper white and alter other colors accordingly. This results in a general decrease in printed luminance for in gamut colors. For instance, if a paper's unprinted white point is L*a*b* 95, -1, -2, a fairly typical glossy white, a L*a*b* value of 80,0, when printed using Relative Colorimetric, will be measured with a spectrophotometer around 76, -1,-2. But when printed in Absolute Colorimetric it will measure as close to 80,0,0 as it's profile's accuracy permits.
When asked to print a color that is out of gamut, and "white" is since papers aren't pure white (100% reflective across the spectrum), the profile will provide a close proximity to that color, usually what is closest as determined by Delta E 1976's metric. Since most papers have a slightly bluish hue, a small amount of yellow ink is added which shifts the b* from a negative value towards 0. Hence the slightly warmer color than the unprinted paper. Since yellow has almost no impact on L* compared to its strongly positive b* shift, this results in a smaller Delta E and is why the print appears warmer.