I can understand your interest in converting LR edits (and presets). While XMP is a fine format for storing certain types of data (e.g., ratings), the interchange ability is not as great as might appear for editing parameters. This is for two main reasons. First, each app has its own implementation of standard constructs (like sharpening, or contrast). Even exposure can be different. Second, apps do not document their algorithms, so app A cannot easily emulate what app B does.
When ACR or Lightroom write out XMP parameters, these are of little use unless the receiving app can convert those values to its own algorithms (and has the algorithms in the first place). Adobe apps can share code and comprehend their own data which is helpful when staying within Adobe’s ecosystem.
So, without algorithm descriptions or code from Adobe, converting adjustments faithfully is pretty impractical. Here is a small snippet from an XMP file from Lightroom:
crs:ProcessVersion=”15.4″
crs:Exposure2012=”-1.61″
crs:Contrast2012=”0″
crs:Highlights2012=”0″
crs:Texture=”-50″
As you can see, the Exposure, Contrast, etc have years attached to them, so they have changed their algorithms over time, and Adobe does not document the algorithms or the range of values. Also, what is the algorithm for “Texture”? Some people have tried to convert these parameters using Machine Learning with mixed success. There is a product called Avalanche that you might look into, though it only handles a subset of LR’s algorithms. In short, I do not have any plans to try to do something similar – it’s far too large of an effort with constant tweaks and infinite bugs.