Time for math! Conserve MP reduces MP cost to 8/16, 9/16, ..., 14/16, 15/16 of the original cost with equal probability (so on average, if it procs, it reduces MP cost to 11.5/16 = 23/32 of the original). Looking at that another way, on average it saves you 9/32 = 28.125% of the original MP cost.
*If* it procs.
Just with the job trait alone, it procs 25% of the time. (And equipment can be applied to any subjob so that's kind of a non-factor.) Which means that it saves you 9/32 * 1/4 = 9/128 = 7.03125% of your *total* MP spending, on average.
So, in order for it to beat out the 20 MP a minute that you get from Auto-Refresh, naturally 7.03125% of your total MP spending per minute (on average), must be greater than 20. Which in turn means that the break-even point is 284.4444 (repeating) total MP spent per minute, on average.
Or, since we're probably talking bigger intervals than single minutes, think of it in, say, 10 minute chunks. In which case Auto-Refresh gives 200 MP, and the break-even point for Conserve MP is 2844.4444 MP spent.
Going with your example, Haste's base recast is 20 seconds, and a reasonable amount of Haste gear bumps that down to more like 16-17, but for the sake of simplicity let's just say you cast three Haste spells a minute, and thus spend 120 MP a minute on Haste. That's way less than the 284 you need to hit the break-even point. And you're not using Haste every 20 seconds constantly: you do a cycle and then rest. So that's even less on average.
You can of course throw in other spells, say you do Haste/Regen 3, three times a minute. Now you've hit 312 MP spent in a minute, and Conserve MP's won out, so your net MP spending is probably more like 290, while Auto-Refresh would've had you at 292. Now can you keep up that level of MP spending every single minute, without resting? Because if you rest for a minute to get all that MP back, that's a minute with 0 MP spent. So the average of those two minutes is now 156, and we're back to "way below 284 MP/minute" again.
In a nutshell, for Conserve MP alone to beat Auto-Refresh, you not only have to have cause to be spending MP constantly, but you also have to have a very hefty amount of MP regeneration just so that you don't eventually run out of MP from all that casting in its own right.
Yes, of course, there will be times when you get lucky with Conserve MP happening unusually often, or happening on particularly high MP cost spells, or reducing MP cost to 8/16 or 9/16 rather than 14/16 or 15/16. There will, statistically, be times when just due to luck, Conserve MP saves you tons of MP.
And then there will be times when it doesn't. And you may not pay as much attention to those. Which is why it's helpful to do the math out.
EDIT: Note that this is not to say that /SMN is unequivocally better for all situations than /BLM. I already said my piece on the comparison of the subs overall. But the situations where you can actually spend enough MP for Conserve MP to beat Auto-Refresh are *very* slim. That doesn't make /BLM useless, because it has a number of utility spells that /SMN lacks, but if you're *strictly* concerned about MP efficiency and not about utility, /SMN almost always wins. Especially if you have to rest or run around (and thus not cast) a lot.