XRF is not better than normal conventional element analysis but much faster and enables you to analyze the elemental range from Na to the  transuranic elements. The d-block transition elements work very well with detection limits in the lower ppm range (based on dry matter).

How reliable are the results from XRF: The answer is quite simple, it depends very much on Your sample prep, a general answer is not possible. To have a start point I would recommend to start  with freeze-dried samples as loose powder and a fundamental parameter based XRF quantification. You could give some of Your samples to an XRF lab to test it if it fits Your needs, analyses are cheap.

Hi Clement, the reliability of XRF depends on several factors. First of all: What is the expected concentration range? (%, ppm, ppb). How do you prepare your specimens? What kind of elements do you want to measure? XRF is only suitable for elements with an order number bigger or equal than 11 (Natrium). Have you a mixture of elements? If yes than you have to take into account that elements in your specimens affect each other. Some elements are excited by the XRF emissions of other elements. With XRF you can analyze solids (e.g. metal discs), powders (e.g. cements) or liquids (here you have to flush measuring chamber with helium because normal atmosphere absorbs to much of your XRF radition). XRF - according to my experience - is suitable for concentrations down to sub-%-range. If you want to analyze in the ppm or ppb-range then ICP-MS or ICP-OES or AAS is better to you, but the costs of ICP are higher than XRF because you need Argon to generate your plasma and Argon is very expensive. Greetings Markus P.S.: Have you ever thought about ion chromatography? It's very sensitive for trace analysis. (ppb range). Well known suppliers are Dionex or Metrohm. Here are some IC resources:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4884567/

http://131.130.57.230/clarotest190/claroline/backends/download.php?url=L01hdGVyaWFsL0lvbnMgaW4gUGxhbnQgZXh0cmFjdHMucGRm&cidReset=true&cidReq=300409WS15

https://watermark.silverchair.com/api/watermark?token=AQECAHi208BE49Ooan9kkhW_Ercy7Dm3ZL_9Cf3qfKAc485ysgAAAiEwggIdBgkqhkiG9w0BBwagggIOMIICCgIBADCCAgMGCSqGSIb3DQEHATAeBglghkgBZQMEAS4wEQQMzW-F1IPju0lRJiWUAgEQgIIB1K3DUiUBk7XhJ7-hobDz35vOnaAc5AqvUFTd1S2eI3NS9-EX9OuxarwQjX_dxm5XKxrgBFzSvfa70WHZeI4sCARhvAS056rFBgsB8CLvFnDerxkgP7UX1BBbwcW8zQ5gXQGM4I2sL_5vwPaPROd2qgSln2JwFgnr0Ddj9AgJw3jkyCFqnlkRvCQdhYD3zEhJopYHjDygCrT6FworY4VBTJ099Ize7ruUHHCjm00mHFGjBjLJ6kJ1ScI9WbYAMPV4WAiGjqTPXersM5fM7lOuUgw8YfJX5DYTkem94ARLLJGymEqVjhNRiBjfLGhXWmGVQgELnvs0w-oKI56WVeiO1Cj5s5I2wypSgOfRTCKKpySqiS_EDT9dqi66AlpLOu0OsHpZrDI517xg5O9bRUrr0bh08YRksDEthjoQCupao5_9y9KbijFciTFbDgKI_CMMqTrwU6Hcixn8dBsWXiA4ih0EIC29UngI0iRZeneMblGcYEEc2pJNRuTwpUTXmoN-dFBJVou5M9F9-OQJLlQT-9MMAn2PzpCjtzuAVfD5J0s27yWesceSHlf15t6iKLaT8mWivHqPpjocfVxtvrLFJMRYVIGQ0XVhWePjVLiXk0upUj5bmg

https://www.researchgate.net/file.PostFileLoader.html?id=594cec1bf7b67e4e47484e22&assetKey=AS%3A508358009593857%401498213402999

Hi Mutsam, I've previously posted your link in my answer but double is better than none :-). Have a  nice day.

The technique is most reliable with a vacuum chamber and wavelength dispersion, especially at low atomic numbers.  At high atomic numbers the signal strength is low, but so is the resolution requirement.  Check out polarised spectroscopy.  Lastly, you need reference samples for calibration.  So it's not worth doing unless you want to analyse a few thousand samples.

XRF is not better than normal conventional element analysis but much faster and enables you to analyze the elemental range from Na to the  transuranic elements. The d-block transition elements work very well with detection limits in the lower ppm range (based on dry matter).

How reliable are the results from XRF: The answer is quite simple, it depends very much on Your sample prep, a general answer is not possible. To have a start point I would recommend to start  with freeze-dried samples as loose powder and a fundamental parameter based XRF quantification. You could give some of Your samples to an XRF lab to test it if it fits Your needs, analyses are cheap.

Not  my skill

To answer the question about repeatability.  For well-prepared plant samples and conditions that give at least thousands of counts above a relatively small background  for an element during the counting period, long-term repeatability is better than that for chemical analysis.  

These references should help answer your questions regarding the reliability and limitations of XRF for plant analysis. 

Hutton, J. T., & Norrish, K. (1977). Plant analyses by X‐ray spectrometry II—Elements of atomic number greater than 20. X‐Ray Spectrometry, 6(1), 12-17.

Norrish, K., & Hutton, J. T. (1977). Plant analyses by X‐ray spectrometry I—Low atomic number elements, sodium to calcium. X‐ray Spectrometry, 6(1), 6-11.

It is very difficult in XRF to duplicate matrices, and yet the matrix radically affects the results. You state that you want a non-destructive method, but sample preparation will render the sample changed enough that you might not be able to use it for other purposes. I'd stick with ICP; you can use low milligram quantities of your sample and it is significantly better for plant concentrations of elements than is XRF.

Quantitative matrix correction is possible but the range of sample matrices must be represented in the calibration and validation sets. Calibration is everything for XRF.