No. That is incorrect. Make a few prints with excellent profiles on various papers ranging from gloss to matte to plain uncoated paper and compare L=50 input to the final printed output values. This test will definitely not yield "constant" mid tone (L=50) results unless using absolute absolute rendering intent or to a lesser extent relative rendering intent without BPC, definitely not with perceptual mapping intents. However, Absolute and Relcol fail miserably for most digital images when printed to matte papers due to severe clipping of the shadow values and out-of-gamut colors among other issues. Hence the need for relcol with BPC and/or perceptual mapping, and hence the variability in mid-point mapping results.
OK, so I did the tests and the results are attached. Minor changes are expected with longer drydown.
CBP2 (Canson Baryta Photographique) shows an L* 50 value of 48, CRP (Canson Rag Photographique Duo) is 52, HPN (Epson Hot Press Natural) is 50. All the preceding were printed in RGB mode, RelCol with BPC (a small uplift is expected with BPC). Also included is HPNBW (Epson Hot Press Natural) in Advanced Black & White mode. While there is some variability around the midpoint, it's close enough that I switch from glossy to matte at will without substantive changes to overall look of the result. In all cases there is a smooth progression to the maximum black for each paper, with compression on the matte papers (as previously discussed). It should be noted that I optimize colour output on gamut, not density. While this test doesn't show what's happening with dark non-neutrals, there's no evidence of any clipping.
The printers used were an Epson 7900 (PK) and 9900 (MK). These results were obtained by printing a 51 stepwedge assigned an L* space, measuring with the output with MeasureTool and processing the results with QuadToneRIP's QTR-Linearize-Data.app.