The other thing that can get you with older CCR tubes is adequate warm up time. Usually 20 minutes is enough. That could explain the shift between the two calibration runs.
If it is a replaceable lamp, you might be able to find a very high CRI equivalent replacement.
If i remember correctly, there is a calibration strip under the starting area of the glass that is used for internal calibration. i don't know of a replacement though if it has yellowed.
That strip is between the glass and the top cover. To replace that in a way that it solves this issue will be hard. In fact that strip is well protected so I doubt it has to be replaced. If you are not using a fluid on the glass for wet mounting and it creeps in between the glass and the strip, like I did :-) If the lamp has to be replaced, a copy with the same spectral output may be wiser, you never know how Epson compensated a lousy spectral output. A good cleaning of the internal glass side was needed on more Epson scanners in time, dust and plasticizer condensation likely causes.
BTW try it too with the target lengthwise scanned and compare it to another direction target profile. The scanner calibration strip is not just setting one value for the strip width but should compensate irregularities over the width of the scan slit. Dust etc accumulated on the glass over the calibration strip spoils that calibration. Dust on the internal mirrors of the optical path to the lens remains the same along the scan path and is compensated in the calibration, dust on top and underneath the scan bed glass during the scan can not be calibrated for.
The lamp and your Wolf Faust target may have degraded in time, the last no longer representing the calibration numbers. AFAIK they were measured per batch and not individually per target. To be honest I have gone that way of scanner calibration more times with either the Wolf Faust iT8 or a CC Passport and it may result with better color for photo dye originals but any other dye or pigment type is not better reproduced. Wonder what the pigment in your monochrome contact prints is, not to mention the paper white. Vuescan probably does a good job for that right away. HP made good use of the flaws in two spectrally different CCFL lamps to nail color. That based on the metameric failure for each. Impossible with one CCFL lamp like in the 3200, the firmware/driver can only correct for a limited set of colorants then, usually the photo dyes.
Using a copy table + four 4000K halogens + a 5DII + the flat field plug-in for Lightroom to even out the lighting more and then creating a RAW color profile with the Passport in that set up is way more rewarding in my experience.
Met vriendelijke groet, Ernst
http://www.pigment-print.com/spectralplots/spectrumviz_1.htmMarch 2017 update, 750+ inkjet media white spectral plots