At the current state of affairs there is a difference between relcol and perceptual, because most profiling sw incorporate a form of BPC in the perceptual tables, which basically means that the blackest black is noted as having an L = 0 value. In such cases, Photoshop will skip its own routines for BPC, even if it's switched on.
BPC is not incorporated in the ICC specs, since it is not an obvious requirement other than for perceptual renderings. And for perceptual intent, the ICC can not make a recommendation, because it is then still open what curve is used for the compensation. A simple linear scaling? Or some form of shadow compression as is the case with Adobe's bpc. In other words, it becomes a judgement call.
And also there is one very important case where BPC is very much NOT a requirement and is to be switched off: namely If you want to make a press proof where the proofing printer blackpoint is darker than the press print blackpoint. In order to faithfully reproduce the lighter black point, you need to switch BPC off.
If you truly want to see the effect of BPC including its shadow compression, you always need to use the Photoshop softproof feature and switch the Black Point Simulation on. This will lighten the blacks if applicable, and allows you to better judge the effect of blacks relative to saturation. If you leave this off, some dark saturated colors may look oversaturated in preview. This is particularly apparent in dark, shadowed skintones...!
Having said all this, I must admit that the first profile I tried from the HM website (FineArt Canvas at the bottom) is indeed influenced by Photoshop BPC, even with Perceptual intent. However, not necessarily in a bad way, and likely because of a less than optimal perceptual table in the profile.
Why is there a difference at all between the two?
Lastly why, since there's an obvious requirement, hasn't this been incorporated into the ICC specs?
Questions, questions ... :-)