Like many other things in the shooting game, high quality calibration weights (meaning ones whose indicated weight values you can trust) are not cheap. If your scale is now off after doing a 2 point calibration, it is likely because of one or both of the following reasons. #1 - One (or more) of your calibration weights are off. #2 - many instruments such as laboratory balances and pH meters can be calibrated using multiple calibration standards (points). The idea is to calibrate the instrument across its entire usable response range. The theory behind this approach is typically that the instrument will have a window in which its response is theoretically linear. However, in many cases the response is not linear. It may be moderately close to linear, but it's not truly linear. Multi-point calibration of such an instrument with high quality, accurate standards will usually result in better accuracy across the usable response range of the instrument. That does not mean it will be dead nuts, spot-on accurate across the entire range, because the response wasn't perfectly linear to begin with. How far each of the calibration standard values fall away the actual response curve of the instrument will give you some idea of how much better the overall calibration across the entire range will be. The farther off each standard falls (i.e. the farther away from true linear response the instruments recorded values are), the farther away from a true (accurate) reading the instrument will give after being calibrated. It may give fairly "good" readings across the entire range, but none of them will be truly "accurate".
Because of reason #2 above, it is not uncommon to see an instrument that will only be used for measurement in a very narrow response window to be calibrated with a single very accurate standard that falls right in the middle of the response window in which it will be used. Most instruments' responses deviate from linear behavior at either end of the usable range. So using calibration standards that fall at either end of the range have a tendency to throw values recorded in the middle of the range farther off from "true". For example, calibrating a pH meter that has an effective usable range of 1.0 to about 12.0 with standards of pH 2.0, 7.0, and 10.0 will typically mean you get pretty good values across the entire range. Not perfect, but pretty good. However, if you're only going to be measuring samples that are all very close to pH 7.0, you might end up with more accurate readings if you use only the single pH 7.0 standard, and don't try to "force" the calibration curve to accommodate the additional calibration points at either end of the range.
Having said all that, how do you "know" which of the weights measured are actually off? If I understand you correctly, you're saying that after you calibrate the scale with both the 20 gr and 50 gr calibration weights, then weigh some of your other calibration weights, the readings you get don't match the value on the weight. So is the scale itself off? Are some of your standard weight values off? Or is the 2-point calibration "forcing" the standard curve to be off? Or some combination of all 3? The easiest fix is simply to go back to a single point calibration with whichever standard weight you were using before. Alternatively, if you know someone with a good analytical balance, you could check your standard weights using a different balance to see what kind of values you get, or even borrow their calibration weights, if they're of good quality. The last solution is to simply buy a quality set yourself. As I think you've already found, they're not cheap.
Your dilemma is that you are now unsure of your powder charge weights due to the calibration issue. However, I can possibly ease your mind somewhat. Many of the things we measure in reloading aren't absolute values. For example, seating depth is usually measured in relation to some point, such as "touching" the lands, if you use something like the Hornady OAL gauge. In reality, the actual "true" measurement could be off by as much as .005", although +/- .002"-.003" should be achievable if you're careful. But it doesn't matter for our purposes because we use it only as a reference point. Our actual seating depths can then be measured very accurately using calipers. So the distance relationship between different seating depths (i.e. - testing in .003" increments) should be very accurate and reproducible, even if they might not represent the "true" distance from the lands. The reason this works is because we ultimately measure the efficacy of changing seating depth on the target. You might believe from your measurements that the optimal seating depth was .015" off the lands, when in reality it was actually .019" off the lands. But it doesn't matter because you can reproduce the the CBTO measurement of that seating depth very accurately with your calipers, any time you load more rounds. So it doesn't matter if it's .015" off or .019" off the lands (your original reference point), you know that rounds with CBTO = 2.641" (just an example) shoot extremely well and you can reproduce rounds with that CBTO measurement whenever you like. The same is true with your balance. If you go back to the single point calibration with the standard weight that worked for you in the past, it is still possible that what you thought all along was 22.50 gr was actually 21.90 gr. As long as your balance reproducibly gives you that weight, you will be fine. It's the difference between precision (reproducibility) and accuracy. Ideally, you want both. But in a case like yours, you will need a high quality known (true) set of standards, and/or access to an additional high quality analytical balance in order to check everything. If your balance was working with a single calibration weight before, it may not be worth the effort and expense for you to try and validate everything.
