Let me look at my own apprentice vocab items and see what single-kanji vocab gets caught in the apprentice-guru-master centrifugal cycle because of mistaken readings… let’s see…
I have six one-word items out of roughly 100 apprentice vocab items. (Review life has been sucky lately.)
陣 （じん）、奥（おく）、価（あたい）、稲（いね）、並（なみ）, 額（がく）…
For me, this problem really only applies to two of these, 稲 and 並. 2%. I have a handle on 稲 but I don’t know how long this will last as the kanji reading is so close (いな). 並 needs more work, and it’s been through the apprentice-guru-master cycle a few times. I know exactly what 並 means and sounds in context or when prompted for vocab meaning.
In Guru I have only two stand-alones out of 110 or so. 銃 is fine. 跡 has this problem… In context I can read it fine, in reviews I have trouble. きみの跡（せき）wouldn’t make sense. きみの跡（あと）does.
Master, out of 110 or so: 旬、甲、鹿、柱、隣、盤、端…
Four of those are affected (bolded). Two of these I would definitely remember if prompted with context or a “vocab” reminder.
There’s not as many items affected by this as I thought, at least not this late in the process. There’s going to be way less single-vocab at level 42 than at level 8, so I barely have to deal with this anymore.
Out of the total number of reviewable items there are not very few single-character vocab items; within the single-character vocab items, there’s a sizeable proportion of items that get sent back due to reading confusion issues. Some leeches fall into this category and get stuck a long time. The inverse of this problem, kanji entered with the vocab meaning, occurs occasionally and is just as annoying, so the total cases of kanji/vocab mixups feels like more; there’s a doubling effect. The scale of the problem could feel bigger because of that.
So there’s some observations and data.
Note: Edited for clarity.