sRGB and aRGB color space both use gamma 2.2 tone ramps which for RGB neutral values of 1,1,1 through 5,5,5, round to L*=1 while RGB values of 6,6,6 through 9,9,9 round to L*=2 in the CIELAB color model (note: RGB 0,0,0 = 0L*). A difference between 1 and 2 L* translates to delta E= 1.0 between the two gray ramps you generated in this deep shadow value. Delta E theory suggests the typical human observer can see delta E =1 increments as "just noticeable", ie. a very small visually discernible difference. At the high end for RGB = 251,251,251 versus 252,252,252, both sets round to L* = 99 (100L = pure diffuse white = RGB 256,256,256). Hence, I think as long as your new profile generates otherwise more well behaved tonal ramps and color fidelity over the full scale, your new custom profile is actually better overall. Even though it loses out in shadow detail by a very tiny amount, the amount is so small that you should have no trouble extracting good shadow detail with appropriate image edits when needed.
I have a monitor checker target that uses the step-off in the low L* values from 0 - 9 and also the highlight L* values 91-100 to help evaluate monitor calibration accuracy. I don't usually print this target since it was designed for verifying monitor performance. That said, if the gray ramps in this target look good on your monitor (ie. your monitor is well calibrated) then a print of this target using appropriate printer profile and Relcol/w BPC might be a good way for you to check how much highlight and shadow detail you are losing when printing straight through the profile with no additional corrections.
Here's the direct link to the MonitorChecker target: http://www.aardenburg-imaging.com/cgi-bin/mrk/_4448ZGxkLzBeMTAwMDAwMDAwMTIzNDU2Nzg5LyoxMDM=