I'd like to take a couple of exceptions to your reasoning here. Second, values varying by "just a little bit" (more than just rounding) have to be taken seriously and accounted for in the formulas. I don't think "fairly close" is good enough for this.
Then firstly, I don't think that's a bug in their table. For instance, if Level=4 & CR=5 was supposed to be what you calculated, then it would be the same number as Level=1-3 & CR=5, which doesn't make sense to me. They may well have bugs in their table, but over the years I've done a good bit of semi-mathematical table analysis on D&D tables, and every time I thought I'd found a typo it turned out to be because I had the wrong formula. So I'd suspect my own formulations first, second, and third before I'd try calling it an error.
I also don't think this should be "fixed", as in coding in an exception for them. This is partially because I hate singular (non-rule-based) exceptions, and partially because I'm not convinced they're wrong.
You might go back to my earlier note in this thread about how most of the table seems to use a sequential 4/3 & 3/2 multiplier pattern, but that pattern breaks down in various places. You'll also note that applying that pattern cumulatively (starting with 3/2) gives the multiplier sequence that you've described, but if you start with 4/3 it gives a completely different sequence in the odd terms, one of which is "5 1/3". My difficulty came when I couldn't figure out why they deviated from a strict pattern.
Also, I'll have to go back and confirm it, but I think I saw this pattern break down in lots of different places near the bottom-left-diagonal edge of the table. This says to me that something else is going on here that we haven't properly accounted for in our pattern. Can you maybe spot any sub-patterns in the list of exceptions?