Yes, I did.
However, it looks like my problem has nothing to do with calibration. When un-calibrated, I still get much better colours (and smoother gradients) in Photoshop when proofing in Monitor RGB than when not proofing.
Calibration makes the colours look more natural (less blueish), but I still have to proof in Monitor RGB to get accurate results.
Therefore, I can only work in the sRGB space.
I realise you've already been through so much - but proofing in the Monitor RGB space is simply turning off screen colour management in Photoshop. You're using the LUT data but nothing else from your profile working this way - and that's not accurate, even for sRGB. Monitor RGB gets you no closer to proper colour management - all you're doing is making Photoshop emulate applications that aren't colour managed. That setting is there for previewing web graphics on your particular monitor, and not much else.
Bearing in mind that you're complaining of a blue tint, your choice of white point may be to blame. 6500K (well, if your calibration solution supports it, D65) is typical, but maybe your monitor, when uncalibrated, is rather warmer than that. This feels unlikely - usually uncalibrated LCD monitors are to the cool side (9300K is not unknown). Nevertheless, it may be worth experimenting with different white points, including profiling to the monitor's 'native' white point if you have that option - much as I'm struggling to see why this would be the issue.
Whilst we're still on the subject of white point, some LCD monitors have an on screen display option to choose the white point. Whilst that sounds like the correct thing to do before calibrating (and is usually the correct thing to do on a CRT, where you're changing the relative gain of the three input amplifiers or similar), it's usually a mistake on an LCD. Leave an LCD monitor set at the native white point, and leave it to the calibration setup to get the white point right. (I've verified this with my setup in Monaco OPTIX XR Pro - the deltaE is worse if I set the white point using the OSD and calibrate than if I leave my panel at its native white point).
A final thing that may be in play is the contrast setting on your LCD if it has one. I drive my monitor using DVI, so the only setting available is brightness which changes the backlight intensity (DVI also gives a sharper picture than using VGA). If you do have a contrast setting on the monitor, try adjusting it before reprofiling - ideally recording the values you've tried.
You haven't said is what monitor you're using - it could be that the monitor itself is bad. A 6 bit TN panel is never going to calibrate well - in fact, no TN panel is that good for serious colour work because of the contrast and colour shifts as you move your head around (which is why laptops are poor for colour critical work). If you have one of those 6 bit abominations that produces 16.2 million colours by frame rate control (check the specification - if it's 16.2 million colours rather than 16.7 million colours, this is probably what you have), that would certainly explain your duff gradients.
I'd argue that gradients are a very poor way to evaluate a profile anyway. A primary colour gradient, say a gradient from R 0 G 0 B 255 to white (R 255 G 255 B 255) will always look smoother in monitor RGB because you're simply varying the R and G values from 0 to 255 along the gradient. With colour management, the blue end of my hypothetical blue gradient is no longer R 0 G 0 B 255 - it's whatever R 0 G 0 B 255 in the working colour space is when transformed to the monitor colour space. The same is true of every intermediate value along the gradient - including white.
R 25 G 13 B 234 doesn't represent a unique colour until you define a colour space. If you say "Monitor RGB", then you get whatever that represents on your monitor with the LUT taken into account. If you say "sRGB" then it represents a particular colour - the colour management system will do some sums to transform R 25 G 13 B 234 in sRGB to the nearest R G B values on your monitor to display the right colour. It could be R 20 G 15 B 225.
Indeed, having the colour management system engaged could explain the colour cast you're observing on white. If you're comparing to white with the LUT loaded but colour management off, that will be R 255 G 255 B 255 transformed by the LUT in the non-colour managed setup. R 255 G 255 B 255 in the colour managed setup goes through a transformation from the white point of the colour space you're using before going through the LUT, and it's possible that the white point of the colour space you're using is out of gamut, hence the apparent colour cast you're seeing. If that's the case, the cast will only affect pure white or very nearly pure white. The in gamut colour chosen by the colour management system for white in the colour space you're working in might not necessarily be neutral.
The banding your image shows is worse than I see on my monitor (Dell 2007WFP with an S-IPS panel, calibrated to D65, 2.2 Gamma with Monaco OPTIX XR Pro in 99 patch table mode), but such banding isn't unknown.
Updating to Photoshop CS2 9.0.2 is a good idea anyway, as there's various bug fixes in 9.0.2 - but I don't think it's going to change anything here.
If you have any non-Adobe colour managed software (such as Canon's Digital Photo Professional if you have a Canon DSLR - though be aware that you have to select the screen profile in DPP as it doesn't pick it up automatically) does that work OK, or is that doing the same as Photoshop?
I strongly suggest you get away from the gradients, and other things that are potentially misleading and are unrepresentative of the colours in your photographs anyway. I suspect your settings are right, though your monitor may be a poor one, making things worse. The Gretag Macbeth calibration setup you're using is a good product - if it was fatally flawed, the Internet would be full of reports about it.
The other thing is, as I already hinted at, you have to give your eyes chance to adjust to working on a calibrated monitor. You get used to the flaws of your monitor when uncalibrated or partially calibrated.
Instead of looking at gradients, look at images like a Gretag Macbeth ColorChecker chart (if you have a ColorChecker and illumination that is 6500K or thereabouts, that will be an even better starting place). Look at photos too - if you have a profiled printer and 'daylight' illumination, how do the colours compare? You could also compare a grey card and mid grey on screen - though be aware that this, too, may look slightly different if your room illumination isn't 6500K or thereabouts. Nevertheless, if you're looking for colour casts on neutral shades, it's best done with mid grey. (Black can be especially misleading on a monitor - IPS monitors can impart something of a violet hue to dark blacks because of the way the screen works - no screen technology is perfect, but of the three main types of LCD screens, my favourite is IPS for photographic work, and I do own examples of all three major types - a TN laptop, a *VA 17 inch LCD and an IPS 20 inch LCD).
In conclusion - what, if any flaws are you seeing in realistic use with colour management enabled, rather than with contrived test cases. Are you seeing banding in skies, for example?