Already 8 years ago (see article "The lab is far from the road," 53x12.com, 2004) I had expressed some doubts about the usefulness of laboratory tests to assess performing capabilities of athletes, not just the elite. 

Even today, scientific publications that evaluate athletes on the field are very few, in cycling even less; I can suggest the following two: 

- "Validation of a field test to Determine the maximal aerobic power in triathletes and endurance cyclists" (Br J Sports Med 2007; 41:174-179), which compares incremental tests on a cyclo-ergometer in the laboratory with similar test (not the same...) carried out on the track, in a velodrome. The study finds a difference in MAP (Maximum Aerobic Performance) of 15% between the test carried out on the track and the test on the ergometer: 407.8w against 354.7w were the average values ​​observed in the 34 athletes participating in the study. 

- "Comparison of a field-based test to estimate functional threshold power and power output at lactate threshold" (J Strength Cond Res 2012; 26:416-421): extrapolates FTP (Functional Threshold Power) from a road test of 8min and compares it with LT4 (threshold at 4 mM/l) measured in the laboratory. Even in this case the compared protocols of effort are not equal. 

The only study, to my knowledge, that compares the same test protocol in the laboratory and on the road is: "Comparacion entre pruebas de laboratorio y campo en ciclistas jovenes de alto nivel para estimacion de umbrales" (Bilbao 28-30 Oct 2010). 
The authors of the Facultad de Ciencias de la Actividad Fisica y del Deporte de la Universidad del Pais Vasco compare the results of the same incremental protocol (35W every 3min) carried out in the laboratory and on the road (8% climb) in nine competing cyclists. 
On average the power at anaerobic threshold (Wobla) in the road test was 10% higher than that measured in the laboratory: 362w against 330w, while the heart rate detected was the same. 
The authors explain the difference in power with the fact that on the road it is possible to stand up on the pedals and therefore use different muscle groups compared to the test in the laboratory. 

In my experience built on thousands of tests on cyclists of all capabilities, I can confirm that the measurements obtained on a simulator (cyclo-ergometer or trainer) underestimate by an average of 10-15% the power outputs detected in similar tests obtained on uphill roads, sometimes with individual differences of up to 25%. 
Therefore, in my opinion, in order to obtain reliable information about training intensities expressed in watts to suggest to cyclists of all levels it is opportune, when possible, to carry out the assessment test on the road, preferably uphill.