I wonder if part of the discrepancy between your target numbers and Hummingbird's is the version of the charts you're looking at. My assumptions are that you've got the final version of the charts while Hummingbird is looking at the (now cleared for open publication? certainly floating around the internet ) version from 15 September 1990, although of course I lack a palantir and cannot tell with certainty what anybody else is looking at.
That September 1990 version, unless I am reading it wrong, shows the same numbers that you listed for 4x4 and GW of 55,620 lb with max A/B per figure 9-2 (Sheet 1 of 12) on page XI-9-3 (note the chart is dated September 1987 despite the first page after the cover being dated September 15th 1990) at SL, 5k, 15k and 25k, but at 35k feet the chart shows Mach 1.97 which is the same number that Hummingbird got.
Nb. I get the same numbers as seen on Figure 9-2 when I cross-reference with Figure 9-5 (5k chart on page XI-9-29, 15k on XI-9-30, 35k on XI-9-31) by tracing a horizontal line at 1g on the Y-axis to its rightmost intersection with the Ps=0 line, and then from that point tracing a vertical line down to the TMN on the X-axis. Not sure if this is the right way to try to cross-reference, and of course Figure 9-5 doesn't have any data for SL or 25k feet.
Another couple of curiosities I see in the September 1990 charts: the 2x2x2 loadout seems to give up basically nothing on the top end compared to 4x4, with a top end speed of maybe Mach 1.94 or 1.95 instead of 1.97 despite being 1500 pounds heavier (cf. XI-9-34), and even slapping on tanks with a bunch of drag and a further gross weight increase of ~2800 lbs still gets you a pretty respectable top end of 1.80 or so - that's actually a fair bit faster than some newer fighters can go when clean, if I'm not mistaken. Also there is a dotted "structural limit" line and a solid "projected structural limit" line at what looks like a +1g offset from the "structural limit" line, for some reason that I am ignorant of.
I wonder whether any of this survived to later versions of the charts based on what I assume is more and better testing.
In any case, thanks for putting in the time and effort dusting off the old charts to ensure that ill-informed and obnoxious pedants such as myself enthusiasts and consumer simmers get a super high-fidelity version of the Tomcat in all its glory.